diff --git a/README.md b/README.md
index d152b38..d5b74ff 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,4 @@
-# zl_bench — FlagGems 单算子性能测试工具
+# zl_bench — FlagGems 算子性能测试工具
 
 针对编译器（FlagTree）改动做算子级 A/B 性能对比的 pytest 插件集 + 驱动脚本。在 FlagGems benchmark 体系之上解决三个问题：
 
@@ -6,6 +6,8 @@
 2. **可解释**：自动按 shape 收集每次运行实际使用的 ttgir（带可读的变体命名），供 IR 级 diff；
 3. **口径统一**：cudagraph 计时消除 launch 开销，小 kernel 的对比不被 CPU 侧噪声淹没；capture 前按 warmup 预算显式预热吸收 autotune/JIT，首轮即稳态、多轮一致。
 
+单算子精测流程之上另有批量筛查层（`run_batch.py`，见「批量多算子测试」），可对双仓库上千个算子做例行体检，产物结构与单算子完全同构。
+
 ## 快速开始
 
 ```bash
@@ -54,6 +56,7 @@ softmax:
 |-----|------|------|
 | `REPLAY_FROM` | 空（record 模式） | 指向某次历史 run 目录，replay 其 autotune 选择。做 A/B 时 B 侧必须设（见「A/B 对比测试标准流程」） |
 | `USE_FLAGTUNE` | `0` | `0` 走普通 autotune（跳过搜索、快速验证）；`1` 用 FlagTune 扩展调优空间（首跑全量搜索、慢）。A/B 两侧须取同值 |
+| `DTYPES` | 空（上游默认 dtype 扫描） | 空格分隔的 dtype 白名单（如 `"bfloat16 float16"`），逐个转为上游 `--dtypes`。算子不支持时上游报 `can't be supported by this op` |
 | `PARALLEL_WARMUP_GPUS` | 空（串行） | 设为 `N`（≥2）时把 autotune sweep 分片到 N 卡并行，测量仍单卡串行；不设或 `<2` 则完全串行（见「加速：多卡并行 sweep」） |
 
 **输出**
@@ -161,11 +164,46 @@ WARNING 1 config key(s) got different winners across shards — parallel interfe
 
 **冲突数就是这次加速的可信度指标**：0 可放心用；偏多说明 config 选择已被污染，出正式结论前不设该变量重跑一遍。其余行为（`REPLAY_FROM` 已设或可见卡不足 2 张时跳过、分片失败回退串行、轮转分片而非按块切、用子进程而非 xdist 的原因）见插件 docstring。分片绑卡遵守继承来的 `CUDA_VISIBLE_DEVICES`——用它选过卡时，分片只会落在你选的那几张上。
 
+## 批量多算子测试（run_batch.py）
+
+单算子流程之上的批量筛查层：按清单逐算子调 `run_pytest.sh`，多卡分片、每卡内部串行，产出逐算子状态表和逐测量行总表。**定位是筛查口径**——跨卡并行测量存在功耗/散热耦合噪声（幅度可到百分之几），发现可疑算子后回单算子串行流程（前几节）确认，不要直接拿批量数字下精细结论。
+
+```bash
+# 全量筛查（双仓库全部可收集算子，排除带无条件 skip 标记的与 ops/dismiss.txt 里的）
+python run_batch.py --all --gpus 0,1,2,3,4,5,6,7 --dtypes bfloat16 --op-timeout 1800
+
+# 指定子集：默认 main 仓，vllm 仓加前缀；同名测试函数用 @文件名 消歧
+python run_batch.py --ops softmax vllm:fused_marlin_moe nextafter_@nextafter_
+
+# 断点续跑（PASS/SKIP 的复用，FAIL/TIMEOUT 的重跑）
+python run_batch.py --all --batch-dir runs/batch_xxx --resume
+
+# 锁 config 复测：逐算子 REPLAY_FROM <root>/<repo>/<op>，语义同单算子 REPLAY_FROM
+python run_batch.py --ops-file my_ops.txt --replay-root runs/batch_xxx
+```
+
+**算子资产（`ops/`）**：
+
+| 文件 | 内容 | 维护方式 |
+|-----|------|---------|
+| `inventory_main.csv` / `inventory_vllm.csv` | 双仓库全部 benchmark 测试函数（op、op_file、无条件 skip 标记、pytest collect 核实结果、op_name 提取） | 上游更新后重跑 `python ops/gen_inventory.py --verify-collect` |
+| `shapes_single.yaml` | 每算子单 shape 的筛查集（537 个精选 op_name 键 + 上游 core_shapes 的类名键底座——上游 shape 回退链是 op_name → MRO 类名 → 基类默认，缺类名键会让 BLAS 族跌到 1 维默认值崩溃） | 手工增补；上游变动后原地刷新：`python ops/gen_inventory.py --migrate-shapes ops/shapes_single.yaml --shapes-out ops/shapes_single.yaml`（未匹配键落 `.unmatched.yaml` 供复核） |
+| `shapes_multi.yaml` | 每算子多 shape 的深查集（结构同上） | 同上 |
+| `dismiss.txt` | 批量排除清单（`[repo:]op` 每行一个），只放本仓复核确认的失败项并注明原因/日期 | 批量跑出 FAIL 并确认原因后手工添加 |
+
+无 shape 条目的算子用上游默认 shape，照常可测。`--shape-file ops/shapes_multi.yaml` 切换深查集。
+
+**执行语义**：每算子独立子进程（CUDA crash 只废单个算子）；`--op-timeout` 对进程组 SIGTERM→SIGKILL，记 `TIMEOUT` 不重试、批次继续；pytest rc=1 与信号杀最多重试 2 次，连续同 rc 视为确定性失败提前止损；`--dtypes` 有两级降级——遇上游 `can't be supported by this op` 即去掉限制重跑（备注 `dtype_fallback`），其余失败且无成功行时也去掉限制最后救一次（备注 `dtype_rescue`，覆盖 torch baseline 对受限 dtype 编译失败的场景）。runtime skipif 与上游 `xfail` 都判 SKIP（上游声明的无信号状态）；`.complete` 标记只给 PASS/SKIP，`--resume` 据此复用。
+
+**产物**：`<batch>/<repo>/<op>/` 与单算子 run 目录结构完全一致（run.log、autotune_records/、ttgir/……），所以任何一个算子都可以事后单独 `REPLAY_FROM` 复测。批量层额外产出 `ops_status.csv`（每算子一行：状态/耗时/行数/备注，边跑边原子更新）与 `summary.csv`（每测量行一行：dtype、latency、speedup、Size Detail，来源是 run.log 的结果表，`batch_summary.py` 也可单独对旧批次重跑）。备注列聚合 `no_cudagraph_rows`/`replay_fallback` 计数，解读口径时先看这列。
+
 ## 计时口径：cudagraph 的预热、回退与精度
 
 **capture 前的显式 warmup**：`do_bench_cudagraph` 自带的内部预热只有 5 次迭代，对首跑要触发 autotune 编译（尤其含 FlagTune 扩展空间）、libtuner 选择、lazy JIT 的算子远远不够——这些一次性开销若漏进被捕获的 graph 或第一个计时迭代，测出的 latency 会 run-to-run 抖动（M=1 多 kernel 路径最明显，实测首轮可低到稳态的 ~1/3）。`_cudagraph_plugin` 因此在 capture 前显式预热：先跑一次并丢弃（吸收 autotune/JIT 编译），再按调用方传入的 warmup 时间预算（`Config.warm_up`）循环稳态预热，然后才 capture。预热次数按稳态单次耗时换算，并夹在 5–200 次之间——亚毫秒 kernel 的实际预热时长因此低于名义预算（1000ms），实测足够；若换新算子仍见首轮抖动，优先调大 `_warmup_before_capture` 里的次数上限。这样首轮即稳态、多轮一致（实测同一 M=1 shape 两轮 speedup 差 <0.1%）。
 
-部分算子本身不支持 CUDA graph capture（测量函数内含 host 同步、动态显存分配、不合法的流操作等），这类算子会自动回退到普通 do_bench 计时，`run.log` 中打 `BENCHMARK_DIRECT_NO_CUDAGRAPH` 标记（含失败阶段与具体原因；`phase=warmup` 表示预热阶段就失败了，并非 graph capture 被拒）。
+部分算子本身不支持 CUDA graph capture（测量函数内含 host 同步、动态显存分配、autograd backward、不合法的流操作等），这类算子会自动回退到普通 do_bench 计时，`run.log` 中打 `BENCHMARK_DIRECT_NO_CUDAGRAPH` 标记（含失败阶段与具体原因；`phase=warmup` 表示预热阶段就失败了，并非 graph capture 被拒）。
+
+回退路径的两个防御措施（都吃过亏）：（1）致命错误判定按错误原文匹配而非宽泛子串——新版 torch 给每个 CUDA 错误都追加 "enable device-side assertions" 提示语，宽松匹配会把所有 capture 失败误判为致命错误、令回退路径整体失效；（2）capture 在一次性 RNG state 替身下执行——`torch.cuda.graph` 会无条件注册默认 CUDA RNG 生成器，capture 中途失败可能让生成器卡在 capturing 态，此后进程内任何 `torch.randn` 都抛 "Offset increment outside graph capture"（毒化显形时机不定，事后修复不可靠），替身隔离让真身永不参与 capture，成败都原样归位。
 
 回退本身不影响 A/B 公平性（两侧同一算子回退行为一致），但**回退口径的测量误差更大**：do_bench 每次迭代都走完整的 Python → launch 路径，kernel 越小，launch 开销和 CPU 侧抖动在数字里占比越高——亚毫秒级 kernel 上两种口径可差 2 倍以上，且行间波动更明显。解读这类算子的结果时：
 
diff --git a/_cudagraph_plugin.py b/_cudagraph_plugin.py
index 3b54a08..9c5564b 100644
--- a/_cudagraph_plugin.py
+++ b/_cudagraph_plugin.py
@@ -8,7 +8,9 @@ fall back to plain do_bench and print a BENCHMARK_DIRECT_NO_CUDAGRAPH marker
 into the log. Before capturing, fn is explicitly warmed up for the caller's
 warmup time budget (do_bench_cudagraph's own 5-iter warmup is too short to
 settle autotune/JIT for the MoE kernels, which makes the first captured
-measurement unstable run-to-run).
+measurement unstable run-to-run). The capture itself runs under a scratch CUDA
+RNG state so that an aborted capture cannot poison the process-wide generator
+(rationale: _capture_with_scratch_rng).
 """
 
 from __future__ import annotations
@@ -25,13 +27,19 @@ _ORIGINAL_DO_BENCH = _tt.do_bench
 def _is_fatal_cuda_error(exc: Exception) -> bool:
     """True for CUDA context errors (illegal address, OOM, device-side assert):
     the context is poisoned, so re-raise rather than fall back and record a bogus
-    latency as success."""
+    latency as success.
+
+    Markers must match the error text itself, not torch's generic footer: newer
+    torch appends "Compile with `TORCH_USE_CUDA_DSA` to enable device-side
+    assertions." to EVERY AcceleratorError, so a loose "device-side assert"
+    marker turned all capture-incompatible ops into fatal re-raises and killed
+    the documented do_bench fallback path entirely."""
     msg = f"{type(exc).__name__}: {exc}".lower()
     markers = (
         "illegal memory access",
         "cudaerrorillegaladdress",
         "out of memory",
-        "device-side assert",
+        "device-side assert triggered",
         "an illegal instruction",
         "misaligned address",
         "uncorrectable ecc",
@@ -83,6 +91,33 @@ def _warmup_before_capture(fn, warmup_ms, grad_to_none):
     torch.cuda.synchronize()
 
 
+def _capture_with_scratch_rng(call):
+    """Run a graph capture under a throwaway CUDA RNG state.
+
+    torch.cuda.graph's capture_begin unconditionally registers the default CUDA
+    RNG generator and sets its capturing flag; when a capture aborts midway the
+    epilogue may never run, the flag sticks, and every later CUDA RNG call in
+    the process (the next shape's torch.randn included) raises "Offset
+    increment outside graph capture". Repairing after the fact is unreliable —
+    when the poisoning surfaces depends on which stage the capture died in —
+    so prevent it instead: swap in a scratch state via graphsafe_set_state
+    before capturing and restore the original state object afterwards, success
+    or failure. The real generator never takes part in a capture, so poisoning
+    can only land on the discarded scratch state, and _seed_plugin determinism
+    is left untouched."""
+    import torch
+    dev = torch.cuda.current_device()
+    gen = torch.cuda.default_generators[dev]
+    saved = gen.graphsafe_get_state()
+    scratch = torch.Generator(device="cuda")
+    scratch.manual_seed(gen.initial_seed())
+    gen.graphsafe_set_state(scratch)
+    try:
+        return call()
+    finally:
+        gen.graphsafe_set_state(saved)
+
+
 def _patched_do_bench(fn, warmup=25, rep=100, grad_to_none=None,
                       quantiles=None, return_mode="mean"):
     """Drop-in replacement for triton.testing.do_bench using cudagraph capture."""
@@ -100,13 +135,13 @@ def _patched_do_bench(fn, warmup=25, rep=100, grad_to_none=None,
         # Settle autotune/JIT before capture (rationale: _warmup_before_capture).
         _warmup_before_capture(fn, warmup, grad_to_none)
         phase = "capture"
-        return _tt.do_bench_cudagraph(
+        return _capture_with_scratch_rng(lambda: _tt.do_bench_cudagraph(
             fn,
             rep=rep,
             grad_to_none=grad_to_none,
             quantiles=quantiles,
             return_mode=return_mode,
-        )
+        ))
     except Exception as exc:
         # Re-raise genuine device failures; anything else falls back to plain
         # do_bench, with the failing phase (sync/warmup/capture) in the marker.
diff --git a/_device_guard_plugin.py b/_device_guard_plugin.py
index f735a1a..bea39a9 100644
--- a/_device_guard_plugin.py
+++ b/_device_guard_plugin.py
@@ -4,7 +4,7 @@ FlagGems' `runtime.backend.device_finder` falls back to
 `subprocess.run(nvidia-smi)` **without a timeout**. Under this conda env's
 forked/inconsistent subprocess (`_posixsubprocess` symbol mismatch), that
 child can hang indefinitely, leaving `import flag_gems` stuck in `wait4`
-(seen as run_pytest "卡住 with no result", GPU 0%).
+(seen as run_pytest hanging with no output, GPU at 0%).
 
 This plugin runs at import time — before any test module imports flag_gems —
 detects the vendor via torch (no subprocess), and sets GEMS_VENDOR so
diff --git a/batch_summary.py b/batch_summary.py
new file mode 100644
index 0000000..3092ecf
--- /dev/null
+++ b/batch_summary.py
@@ -0,0 +1,186 @@
+#!/usr/bin/env python3
+"""聚合一次批量跑（run_batch.py 产物目录）的结果为 summary.csv。
+
+数据源是各算子 run 目录里的 run.log（run_pytest.sh 已去 ANSI 色）：
+- 逐表解析 `Operator: <op_name> (dtype=..., mode=..., level=...)` 标题
+  （兼容 _pretty_report_plugin 与上游原生两种格式）；
+- 逐行解析 SUCCESS/FAILED 表行：状态 + 前三个数值列固定为
+  torch_lat / gems_lat / speedup（tflops/gbps 等追加列收进 extra_metrics），
+  余下为 size_detail；
+- 汇入 run_batch.py 写的 ops_status.csv（每算子总状态 / 备注）。
+
+也可单独使用：python batch_summary.py runs/batch_xxx [-o summary.csv]
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import os
+import re
+import sys
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+# 兼容两种标题：
+#   pretty:   Operator: softmax  (dtype=torch.float16, mode=kernel, level=core)
+#   upstream: Operator: softmax  Performance Test (dtype=torch.float16, mode=kernel,level=core)
+OPERATOR_HEADER_RE = re.compile(
+    r"^Operator:\s+(?P<name>\S+)\s+(?:Performance Test\s*)?"
+    r"\(dtype=(?P<dtype>[^,]+),\s*mode=(?P<mode>[^,]+),\s*level=(?P<level>[^)]+)\)"
+)
+ROW_RE = re.compile(r"^(?P<status>SUCCESS|FAILED)\s+(?P<rest>\S.*)$")
+NUM_TOKEN_RE = re.compile(
+    r"^(?:N/A|nan|-?inf"
+    r"|-?(?:\d+(?:\.\d*)?|\.\d+)(?:[eE][+-]?\d+)?)$"
+)
+PYTEST_SKIPPED_RE = re.compile(r"(?P<n>\d+)\s+(?:skipped|xfailed)\b")
+DIRECT_NO_CUDAGRAPH_RE = re.compile(r"BENCHMARK_DIRECT_NO_CUDAGRAPH\b")
+REPLAY_FALLBACK_RE = re.compile(r"AUTOTUNE_REPLAY_FALLBACK\b")
+
+SUMMARY_FIELDS = [
+    "repo", "op", "fg_op_name", "dtype", "case_id", "row_status",
+    "torch_lat", "gems_lat", "speedup", "extra_metrics", "size_detail",
+    "op_status", "notes",
+]
+
+
+def parse_run_log(text: str) -> List[Dict[str, str]]:
+    """把 run.log 里的全部结果表解析为行列表（跨多个 dtype 表）。"""
+    rows: List[Dict[str, str]] = []
+    cur_op, cur_dtype = "", ""
+    for line in text.splitlines():
+        stripped = line.strip()
+        header = OPERATOR_HEADER_RE.match(stripped)
+        if header:
+            cur_op = header.group("name")
+            cur_dtype = header.group("dtype").strip()
+            continue
+        m = ROW_RE.match(stripped)
+        if not m:
+            continue
+        tokens = m.group("rest").split()
+        metrics: List[str] = []
+        while tokens and len(metrics) < 7 and NUM_TOKEN_RE.match(tokens[0]):
+            metrics.append(tokens.pop(0))
+        # 前三个数值列固定：torch_lat, gems_lat, speedup（上游列序承诺不变）
+        while len(metrics) < 3:
+            metrics.append("")
+        rows.append({
+            "fg_op_name": cur_op,
+            "dtype": cur_dtype,
+            "row_status": m.group("status"),
+            "torch_lat": metrics[0],
+            "gems_lat": metrics[1],
+            "speedup": metrics[2],
+            "extra_metrics": " ".join(metrics[3:]),
+            "size_detail": " ".join(tokens),
+        })
+    return rows
+
+
+def pytest_skipped_count(text: str) -> int:
+    """runtime skipif 与上游 xfail 标记都算"无测量信号"，供 SKIP 判定。"""
+    count = 0
+    for m in PYTEST_SKIPPED_RE.finditer(text[-4000:]):
+        count += int(m.group("n"))
+    return count
+
+
+def marker_notes(text: str) -> str:
+    """cudagraph 回退 / replay 兜底标记计数，供备注列（口径解读用）。"""
+    notes = []
+    n = len(DIRECT_NO_CUDAGRAPH_RE.findall(text))
+    if n:
+        notes.append(f"no_cudagraph_rows={n}")
+    n = len(REPLAY_FALLBACK_RE.findall(text))
+    if n:
+        notes.append(f"replay_fallback={n}")
+    return ";".join(notes)
+
+
+def classify(rc: int, rows: List[Dict[str, str]], log_text: str) -> str:
+    """算子级状态：FAIL / SKIP / PASS（TIMEOUT 由 run_batch 在外层判）。"""
+    if rc != 0:
+        return "FAIL"
+    if not rows:
+        return "SKIP" if pytest_skipped_count(log_text) else "FAIL"
+    if any(r["row_status"] != "SUCCESS" for r in rows):
+        return "FAIL"
+    return "PASS"
+
+
+def _load_ops_status(batch_dir: Path) -> Dict[str, Dict[str, str]]:
+    path = batch_dir / "ops_status.csv"
+    if not path.is_file():
+        return {}
+    with path.open(newline="") as f:
+        return {f"{r['repo']}/{r['run_name']}": r for r in csv.DictReader(f)}
+
+
+def summarize(batch_dir: Path, out_path: Optional[Path] = None) -> Path:
+    batch_dir = batch_dir.resolve()
+    out_path = out_path or (batch_dir / "summary.csv")
+    status_by_dir = _load_ops_status(batch_dir)
+
+    all_rows: List[Dict[str, str]] = []
+    for repo_dir in sorted(p for p in batch_dir.iterdir() if p.is_dir()):
+        repo = repo_dir.name
+        for op_dir in sorted(p for p in repo_dir.iterdir() if p.is_dir()):
+            log = op_dir / "run.log"
+            if not log.is_file():
+                continue
+            text = log.read_text(errors="replace")
+            status_row = status_by_dir.get(f"{repo}/{op_dir.name}", {})
+            op = status_row.get("op") or op_dir.name
+            table_rows = parse_run_log(text)
+            if not table_rows:
+                all_rows.append({
+                    "repo": repo, "op": op, "fg_op_name": "", "dtype": "",
+                    "case_id": "", "row_status": "", "torch_lat": "",
+                    "gems_lat": "", "speedup": "", "extra_metrics": "",
+                    "size_detail": "",
+                    "op_status": status_row.get("status", ""),
+                    "notes": status_row.get("notes", ""),
+                })
+                continue
+            case_counter: Dict[Tuple[str, str], int] = {}
+            for row in table_rows:
+                key = (row["fg_op_name"], row["dtype"])
+                idx = case_counter.get(key, 0)
+                case_counter[key] = idx + 1
+                all_rows.append({
+                    "repo": repo, "op": op, "case_id": f"case_{idx:03d}",
+                    "op_status": status_row.get("status", ""),
+                    "notes": status_row.get("notes", ""),
+                    **row,
+                })
+
+    tmp = out_path.with_name(out_path.name + ".tmp")
+    with tmp.open("w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=SUMMARY_FIELDS)
+        writer.writeheader()
+        writer.writerows(all_rows)
+    os.replace(tmp, out_path)
+    return out_path
+
+
+def main() -> None:
+    p = argparse.ArgumentParser(description=__doc__.splitlines()[0])
+    p.add_argument("batch_dir", type=Path, help="run_batch.py 的产物目录")
+    p.add_argument("-o", "--output", type=Path, default=None,
+                   help="输出 CSV（默认 <batch_dir>/summary.csv）")
+    args = p.parse_args()
+    if not args.batch_dir.is_dir():
+        raise SystemExit(f"目录不存在: {args.batch_dir}")
+    out = summarize(args.batch_dir, args.output)
+    with out.open(newline="") as f:
+        rows = list(csv.DictReader(f))
+    n_ok = sum(1 for r in rows if r["row_status"] == "SUCCESS")
+    n_bad = sum(1 for r in rows if r["row_status"] == "FAILED")
+    print(f"[summary] {out}: {len(rows)} 行（SUCCESS={n_ok} FAILED={n_bad}）",
+          file=sys.stderr)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/ops/dismiss.txt b/ops/dismiss.txt
new file mode 100644
index 0000000..2142d4b
--- /dev/null
+++ b/ops/dismiss.txt
@@ -0,0 +1,120 @@
+# 批量测试排除清单。每行一个算子：`op`（默认 main 仓）或 `repo:op`。
+# run_batch.py --all 会跳过这里列出的算子；带无条件 pytest.mark.skip 的算子由
+# inventory 的 marked_skip 列自动剔除，不必写在这里。
+# 只放"经本仓批量跑复核确认失败"的条目，注明原因与复核日期。
+#
+# 2026-08-12 复核记录：
+# - 旧回归清单 24 个候选全部重验：14 个新上游已修复（移出），3 个系本仓
+#   _cudagraph_plugin 回退失效 bug（已修复，移出），7 个确认为上游/环境问题；
+# - 全量筛查（1103 算子）新增确认 67 个：43 个 torch jiterator 环境问题 +
+#   24 个上游 bug/环境限制。shape 语义类失败（28 个）已通过 shape yaml 合并
+#   上游 core_shapes 类名键修复，不在此列；
+# - test_FLA/ 补齐后新增确认 2 个（硬件限制 + 专用 harness），共 76 条。
+
+# ==== 上游 kernel 编译错误 [2026-08-12]
+# CompilationError（关 cudagraph 后仍复现）
+hstack
+scaled_mm_benchmark
+scaled_mm_out_benchmark
+# tt.load result/ptr type 校验失败
+flash_mla_with_kvcache
+# PassManager::run failed（FlagTree triton 编译失败）
+flash_mla_sparse_fwd
+vllm:flash_mla_sparse_fwd
+# capture 失败回退后仍报 kernel 错误
+vllm:flash_mla
+vllm:flash_mla_with_kvcache
+linalg_lu_factor_ex
+
+# ==== 上游 benchmark 代码 bug [2026-08-12]
+# baseline torch.trace 收到非矩阵输入（expected a matrix）
+trace
+# set_shapes() 签名与基类调用不符（TypeError）
+get_paged_mqa_logits_metadata
+# 输入解包 ValueError（expected 3, got 1）
+grouped_mm
+# 上游 benchmark assert 失败
+get_scheduler_metadata
+# is_backward 对未参与计算图的张量求梯度（allow_unused）
+embedding_backward
+margin_ranking_loss_backward
+# baseline 函数签名不符（unexpected keyword 'q' / 缺 constexpr 实参）
+fp8_fp4_paged_mqa_logits
+vllm:fp8_fp4_mqa_logits
+# baseline 返回 None（NoneType has no attribute 'device'）
+special_gammainc
+# normalized_shape=[]（上游默认 shape 语义错误）
+native_layer_norm
+# 上游断言 K>=16 与其默认 attention shape 冲突
+perf_scaled_dot_product_flash_attention_backward
+perf_scaled_dot_product_cudnn_attention_backward
+# Triton Error [CUDA]: invalid argument（上游默认 shape 下 kernel 启动参数非法）
+index_select_backward
+pairwise_distance
+# test_blas_perf_parallel 专用 harness（shape kind 语义与通用注入冲突；常规 mul 已覆盖）
+perf_mul
+# 需要 pytest-benchmark fixture（非 FlagGems Benchmark 体系）
+vllm:triton_unified_attention_perf
+# 自定义输出格式（非 FlagGems Benchmark 表）且预编译超 40min
+vllm:perf_chunk_gla
+
+# ==== 硬件限制 [2026-08-12]
+# kernel 共享内存需求 245-335KB 超 H20 上限 227KB，autotune 全候选 OutOfResources
+vllm:chunk_gdn2
+
+# ==== 环境限制（换环境后应重验）[2026-08-12]
+# vllm 0.20.2 无 vllm.utils.deep_gemm.fp8_mqa_logits
+fp8_mqa_logits
+# FA2 不支持 num_splits > 1（flash-attn 版本）
+flash_attn_varlen_opt_func
+# magma 显存分配失败（cannot allocate memory on GPU, info=-113）
+cholesky_solve
+cholesky_solve_out
+# cusolver INTERNAL_ERROR（Xgeev）
+linalg_eigvals
+# ---- torch jiterator/NVRTC 在本环境编译失败（fp32 也复现），
+# ---- sinc/bessel/special 族的 torch baseline 整体不可用：
+erfc
+erfc_
+erfinv
+erfinv_
+lcm
+lcm_
+lgamma
+lgamma_
+mvlgamma
+mvlgamma_
+polygamma
+polygamma_inplace
+polygamma_out
+sinc
+sinc_
+special_airy_ai
+special_airy_ai_out
+special_bessel_j0
+special_bessel_j1
+special_chebyshev_polynomial_u
+special_chebyshev_polynomial_v
+special_chebyshev_polynomial_w
+special_chebyshev_polynomial_w_out
+special_erfc
+special_erfcx
+special_erfinv
+special_erfinv_out
+special_gammaln
+special_gammaln_out
+special_hermite_polynomial_h
+special_i1e
+special_legendre_polynomial_p
+special_log1p_non_tensor
+special_modified_bessel_k0
+special_modified_bessel_k0_out
+special_modified_bessel_k1
+special_modified_bessel_k1_out
+special_scaled_modified_bessel_k1
+special_scaled_modified_bessel_k1_out
+special_shifted_chebyshev_polynomial_u
+special_shifted_chebyshev_polynomial_u_
+special_shifted_chebyshev_polynomial_v
+special_shifted_chebyshev_polynomial_w
+special_sinc
diff --git a/ops/gen_inventory.py b/ops/gen_inventory.py
new file mode 100644
index 0000000..b2dc9e7
--- /dev/null
+++ b/ops/gen_inventory.py
@@ -0,0 +1,294 @@
+#!/usr/bin/env python3
+"""重建批量测试的算子清单（inventory），并可校验迁移 shape yaml。
+
+对每个 FlagGems 仓库静态扫描 benchmark/ 下的 test_*.py（含 test_ 前缀的
+子目录，如 test_FLA/——op_file 记相对路径去前后缀，run_pytest.sh 的
+`benchmark/test_${OP_FILE}.py` 拼接对其天然成立；models_benchmark/ 等
+非 test_ 前缀路径不在性能批量范围）：
+
+- 每个模块级 test_* 函数出一行 inventory：op（函数名去 test_ 前缀，即
+  run_pytest.sh 的 OP）、op_file（文件名去 test_ 前缀/.py 后缀，即 OP_FILE）；
+- 标注 marked_skip（函数带无条件 pytest.mark.skip，跑了必 SKIP，批量驱动默认剔除）；
+- 提取 op_name 字符串常量（op_name= 关键字实参，以及 *Benchmark 类
+  __init__ 里 super().__init__ 的首个字符串位置实参），用于 shape yaml 键校验；
+- --verify-collect 时额外跑一次 pytest --collect-only 核实函数确实可被收集
+  （AST 见到 ≠ pytest 收得到，import 失败/条件定义都会导致差异）。
+
+产物：
+  ops/inventory_<label>.csv        # repo,op,op_file,file,marked_skip,collected,op_names
+  --migrate-shapes 时按 op_name 并集过滤输入 yaml、合并上游 core_shapes 底座：
+  <out>.yaml + <out>.unmatched.yaml（未匹配任何 op_name 的键，供人工复核后删除）
+
+用法示例（默认扫描两个新仓库）：
+  python ops/gen_inventory.py --verify-collect
+  # 上游 op_name/core_shapes 变动后，对现有 shape 集重新校验（输入=输出即原地刷新）
+  python ops/gen_inventory.py --migrate-shapes ops/shapes_single.yaml \
+      --shapes-out ops/shapes_single.yaml
+"""
+
+from __future__ import annotations
+
+import argparse
+import ast
+import csv
+import os
+import subprocess
+import sys
+from pathlib import Path
+from typing import Dict, List, Optional, Set, Tuple
+
+import yaml
+
+OPS_DIR = Path(__file__).resolve().parent
+
+DEFAULT_REPOS = [
+    ("main", Path("/workspace/dev/FlagGems")),
+    ("vllm", Path("/workspace/dev/FlagGems-vllm")),
+]
+
+INVENTORY_FIELDS = ["repo", "op", "op_file", "file", "marked_skip", "collected", "op_names"]
+
+
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(description=__doc__.splitlines()[0])
+    p.add_argument("--repo", action="append", default=None, metavar="LABEL=PATH",
+                   help="要扫描的仓库（可多次），默认 main=/workspace/dev/FlagGems "
+                        "与 vllm=/workspace/dev/FlagGems-vllm")
+    p.add_argument("--verify-collect", action="store_true",
+                   help="用 pytest --collect-only 核实每个测试函数可被收集（慢，"
+                        "需要完整运行环境；结果写进 collected 列）")
+    p.add_argument("--out-dir", type=Path, default=OPS_DIR,
+                   help="inventory CSV 输出目录（默认 ops/）")
+    p.add_argument("--migrate-shapes", type=Path, default=None,
+                   help="旧 shape yaml 路径；按扫描出的 op_name 并集过滤后迁移")
+    p.add_argument("--shapes-out", type=Path, default=None,
+                   help="迁移后 yaml 输出路径（--migrate-shapes 时必填）")
+    return p.parse_args()
+
+
+def _resolve_repos(raw: Optional[List[str]]) -> List[Tuple[str, Path]]:
+    if not raw:
+        return [(label, path) for label, path in DEFAULT_REPOS if path.is_dir()]
+    repos = []
+    for item in raw:
+        label, _, path = item.partition("=")
+        if not path:
+            raise SystemExit(f"--repo 需要 LABEL=PATH 形式，收到: {item!r}")
+        repos.append((label, Path(path)))
+    return repos
+
+
+def _is_unconditional_skip(dec: ast.AST) -> bool:
+    """pytest.mark.skip（非 skipif）——带不带 reason 实参都算无条件。"""
+    target = dec.func if isinstance(dec, ast.Call) else dec
+    if not (isinstance(target, ast.Attribute) and target.attr == "skip"):
+        return False
+    mark = target.value
+    return (isinstance(mark, ast.Attribute) and mark.attr == "mark"
+            and isinstance(mark.value, ast.Name) and mark.value.id == "pytest")
+
+
+def _op_names_in(node: ast.AST) -> Set[str]:
+    """节点范围内的 op_name 字符串常量，覆盖三种上游写法：
+    1. 任意调用的 op_name= 关键字实参；
+    2. super().__init__ 的首个字符串位置实参（FusedDeepseekV4... 模式）；
+    3. *Benchmark 类实例化的首个字符串位置实参（ScaledMMBenchmark("scaled_mm",
+       ...) 这类经形参转发、常量提取够不到 super() 调用的模式）。
+    误报只会让 shape yaml 多留一个无人读取的键，无害。"""
+    names: Set[str] = set()
+    for sub in ast.walk(node):
+        if not isinstance(sub, ast.Call):
+            continue
+        for kw in sub.keywords:
+            if kw.arg == "op_name" and isinstance(kw.value, ast.Constant) \
+                    and isinstance(kw.value.value, str):
+                names.add(kw.value.value)
+        first_str = (sub.args[0].value
+                     if sub.args and isinstance(sub.args[0], ast.Constant)
+                     and isinstance(sub.args[0].value, str) else None)
+        if first_str is None:
+            continue
+        func = sub.func
+        if (isinstance(func, ast.Attribute) and func.attr == "__init__"
+                and isinstance(func.value, ast.Call)
+                and isinstance(func.value.func, ast.Name)
+                and func.value.func.id == "super"):
+            names.add(first_str)
+        callee = func.id if isinstance(func, ast.Name) else (
+            func.attr if isinstance(func, ast.Attribute) else "")
+        if callee.endswith("Benchmark"):
+            names.add(first_str)
+    return names
+
+
+def scan_repo(label: str, root: Path) -> Tuple[List[Dict[str, str]], Set[str]]:
+    """返回 (inventory 行, 该仓库全部 op_name 集合)。"""
+    bench = root / "benchmark"
+    rows: List[Dict[str, str]] = []
+    op_names_all: Set[str] = set()
+    for path in sorted(bench.rglob("test_*.py")):
+        rel = path.relative_to(bench)
+        # 仅收 test_ 前缀链路（根级 test_x.py 与 test_XXX/ 子目录），保证
+        # "benchmark/test_" + op_file + ".py" 能原样重建路径。
+        if not str(rel).startswith("test_"):
+            continue
+        try:
+            tree = ast.parse(path.read_text())
+        except SyntaxError as exc:
+            print(f"[gen] warning: {path} 解析失败，跳过: {exc}", file=sys.stderr)
+            continue
+        op_names_all |= _op_names_in(tree)
+        op_file = str(rel)[len("test_"):-len(".py")]
+        for node in tree.body:
+            if not isinstance(node, (ast.FunctionDef, ast.AsyncFunctionDef)):
+                continue
+            if not node.name.startswith("test_"):
+                continue
+            marked = any(_is_unconditional_skip(d) for d in node.decorator_list)
+            func_ops = sorted(_op_names_in(node))
+            rows.append({
+                "repo": label,
+                "op": node.name[len("test_"):],
+                "op_file": op_file,
+                "file": f"benchmark/{rel}",
+                "marked_skip": "yes" if marked else "",
+                "collected": "",
+                "op_names": ";".join(func_ops),
+            })
+    return rows, op_names_all
+
+
+def verify_collect(label: str, root: Path, rows: List[Dict[str, str]]) -> None:
+    """跑 pytest --collect-only 核实测试函数可被收集，写 collected 列。"""
+    env = os.environ.copy()
+    env.setdefault("GEMS_VENDOR", "nvidia")  # 跳过 nvidia-smi 子进程探测（可能挂死）
+    proc = subprocess.run(
+        [sys.executable, "-m", "pytest", "--collect-only", "-q",
+         "--continue-on-collection-errors", "benchmark/"],
+        cwd=str(root), text=True, capture_output=True, env=env, timeout=1200,
+    )
+    collected_funcs: Set[Tuple[str, str]] = set()
+    for line in proc.stdout.splitlines():
+        line = line.strip()
+        if "::" not in line or not line.startswith("benchmark/"):
+            continue
+        file_part, _, rest = line.partition("::")
+        func = rest.split("::")[0].split("[")[0]
+        collected_funcs.add((file_part, func))
+    if not collected_funcs:
+        print(f"[gen] warning: {label} collect 无结果 (rc={proc.returncode})，"
+              f"stderr 尾部: {proc.stderr[-500:]}", file=sys.stderr)
+        return
+    for row in rows:
+        key = (row["file"], f"test_{row['op']}")
+        row["collected"] = "yes" if key in collected_funcs else "no"
+    n_missing = sum(1 for r in rows if r["collected"] == "no")
+    if n_missing:
+        print(f"[gen] {label}: {n_missing} 个 AST 可见但 pytest 未收集到的函数"
+              f"（import 失败/条件定义），inventory 中 collected=no")
+
+
+def carry_over_collected(out_dir: Path, label: str,
+                         rows: List[Dict[str, str]]) -> None:
+    """未跑 --verify-collect 时，从已有 CSV 继承 collected 列，避免重写清单时
+    把上次核实的结果冲掉。"""
+    path = out_dir / f"inventory_{label}.csv"
+    if not path.is_file():
+        return
+    with path.open(newline="") as f:
+        prev = {(r["op"], r["op_file"]): r.get("collected", "")
+                for r in csv.DictReader(f)}
+    for row in rows:
+        row["collected"] = prev.get((row["op"], row["op_file"]), "")
+
+
+def write_inventory(out_dir: Path, label: str, rows: List[Dict[str, str]]) -> Path:
+    out = out_dir / f"inventory_{label}.csv"
+    tmp = out.with_name(out.name + ".tmp")
+    with tmp.open("w", newline="") as f:
+        writer = csv.DictWriter(f, fieldnames=INVENTORY_FIELDS)
+        writer.writeheader()
+        writer.writerows(rows)
+    os.replace(tmp, out)
+    return out
+
+
+def migrate_shapes(old_path: Path, out_path: Path, valid_op_names: Set[str],
+                   repos: List[Tuple[str, Path]]) -> None:
+    """迁移旧 shape yaml，并以上游 core_shapes.yaml 为底座合并。
+
+    上游 set_shapes 的回退链是 op_name 键 → MRO 类名键（BlasBenchmark 等）→
+    基类 DEFAULT_SHAPES（1 维）。--shape_file 是整体替换而非叠加，若我们的
+    yaml 缺少类名级条目，BLAS 这类需要 (B,M,N,K) 的算子会跌到 1 维默认值上
+    解包崩溃。因此把各仓库 core_shapes.yaml 里我们没有的键（含类名键与
+    op 级键）原样并入，迁移条目优先，先并 main 后并 vllm（同名键先到先得）。"""
+    data = yaml.safe_load(old_path.read_text()) or {}
+    core_maps = []
+    for label, root in repos:
+        core = root / "benchmark" / "core_shapes.yaml"
+        if core.is_file():
+            core_maps.append((label, yaml.safe_load(core.read_text()) or {}))
+    core_keys = {k for _, m in core_maps for k in m}
+    kept = {k: v for k, v in data.items() if k in valid_op_names}
+    # core_shapes 来源的键（类名键等）不算"未匹配"——它们每次都从上游取新版，
+    # 这样输入=输出的在地刷新也能同步上游 core_shapes 的变动。
+    dropped = {k: v for k, v in data.items()
+               if k not in valid_op_names and k not in core_keys}
+    n_curated = len(kept)
+    merged_from = []
+    for label, m in core_maps:
+        n_before = len(kept)
+        for key, val in m.items():
+            kept.setdefault(key, val)
+        merged_from.append(f"{label}:+{len(kept) - n_before}")
+    header = (f"# 由 gen_inventory.py 从 {old_path.name} 迁移：保留新上游仍存在的"
+              f" op_name 键 {n_curated}/{len(data)}，"
+              f"并合并上游 core_shapes.yaml 缺失键（{' '.join(merged_from)}）。\n"
+              "# 顶层键 = FlagGems op_name 或 Benchmark 类名（上游回退链需要）；\n"
+              "# 两者都无键的算子使用其基类 DEFAULT_SHAPES。\n")
+    out_path.write_text(header + yaml.safe_dump(
+        kept, sort_keys=True, default_flow_style=None, allow_unicode=True))
+    if dropped:
+        unmatched = out_path.with_suffix(".unmatched.yaml")
+        unmatched.write_text(
+            f"# {old_path.name} 中未匹配新上游任何 op_name 的键（{len(dropped)} 个），"
+            "供人工复核后手动挪回。\n"
+            + yaml.safe_dump(dropped, sort_keys=True, default_flow_style=None,
+                             allow_unicode=True))
+        print(f"[gen] shapes: 保留 {len(kept)}，剔除 {len(dropped)} -> {unmatched}")
+    else:
+        print(f"[gen] shapes: 全部 {len(kept)} 个键有效")
+    # 覆盖率：有效 op_name 里有多少没有 shape 条目（用上游默认 shape，仅提示）
+    uncovered = sorted(valid_op_names - set(kept))
+    print(f"[gen] shapes: 新上游 {len(valid_op_names)} 个 op_name 中 "
+          f"{len(uncovered)} 个无 shape 条目（将用上游默认 shape）")
+
+
+def main() -> None:
+    args = parse_args()
+    repos = _resolve_repos(args.repo)
+    if not repos:
+        raise SystemExit("没有可扫描的仓库")
+    args.out_dir.mkdir(parents=True, exist_ok=True)
+
+    all_op_names: Set[str] = set()
+    for label, root in repos:
+        rows, op_names = scan_repo(label, root)
+        all_op_names |= op_names
+        if args.verify_collect:
+            verify_collect(label, root, rows)
+        else:
+            carry_over_collected(args.out_dir, label, rows)
+        out = write_inventory(args.out_dir, label, rows)
+        n_skip = sum(1 for r in rows if r["marked_skip"])
+        print(f"[gen] {label}: {len(rows)} 个测试函数 -> {out}"
+              f"（其中 {n_skip} 个带无条件 skip 标记）")
+
+    if args.migrate_shapes:
+        if not args.shapes_out:
+            raise SystemExit("--migrate-shapes 需要同时给 --shapes-out")
+        migrate_shapes(args.migrate_shapes, args.shapes_out, all_op_names, repos)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/ops/inventory_main.csv b/ops/inventory_main.csv
new file mode 100644
index 0000000..2e714a7
--- /dev/null
+++ b/ops/inventory_main.csv
@@ -0,0 +1,1016 @@
+repo,op,op_file,file,marked_skip,collected,op_names
+main,Concat,Concat,benchmark/test_Concat.py,,yes,concat
+main,perf_chunk_gated_delta_rule_fwd,FLA/test_chunk_gated_delta_rule_fwd,benchmark/test_FLA/test_chunk_gated_delta_rule_fwd.py,,yes,chunk_gated_delta_rule_fwd
+main,perf_chunk_gated_delta_rule,FLA/test_chunk_gated_delta_rule_perf,benchmark/test_FLA/test_chunk_gated_delta_rule_perf.py,,yes,chunk_gated_delta_rule
+main,perf_fused_recurrent_gated_delta_rule,FLA/test_fused_recurrent_gated_delta_rule_perf,benchmark/test_FLA/test_fused_recurrent_gated_delta_rule_perf.py,,yes,fused_recurrent_gated_delta_rule
+main,_reshape_alias,_reshape_alias,benchmark/test__reshape_alias.py,,yes,_reshape_alias
+main,abs,abs,benchmark/test_abs.py,,yes,abs
+main,abs_inplace,abs,benchmark/test_abs.py,,yes,abs_
+main,absolute,absolute,benchmark/test_absolute.py,,yes,absolute
+main,acos,acos,benchmark/test_acos.py,,yes,acos
+main,acosh,acosh,benchmark/test_acosh.py,,yes,acosh
+main,acosh_,acosh,benchmark/test_acosh.py,,yes,acosh_
+main,act_quant_perf,act_quant,benchmark/test_act_quant.py,,yes,act_quant_triton
+main,perf_adaptive_avg_pool2d,adaptive_avg_pool2d,benchmark/test_adaptive_avg_pool2d.py,,yes,adaptive_avg_pool2d
+main,adaptive_avg_pool2d_backward,adaptive_avg_pool2d_backward,benchmark/test_adaptive_avg_pool2d_backward.py,,yes,adaptive_avg_pool2d_backward
+main,adaptive_max_pool2d_backward,adaptive_max_pool2d_backward,benchmark/test_adaptive_max_pool2d_backward.py,,yes,adaptive_max_pool2d_backward
+main,adaptive_max_pool3d_backward,adaptive_max_pool3d_backward,benchmark/test_adaptive_max_pool3d_backward.py,,yes,adaptive_max_pool3d_backward
+main,add,add,benchmark/test_add.py,,yes,add
+main,add_inplace,add,benchmark/test_add.py,,yes,add_
+main,add_relu,add_relu,benchmark/test_add_relu.py,,yes,add_relu
+main,add_rms_norm,add_rms_norm,benchmark/test_add_rms_norm.py,,yes,add_rms_norm
+main,addbmm,addbmm,benchmark/test_addbmm.py,,yes,addbmm
+main,addcdiv,addcdiv,benchmark/test_addcdiv.py,,yes,addcdiv
+main,addcdiv_out,addcdiv,benchmark/test_addcdiv.py,,yes,addcdiv_out
+main,addcdiv_,addcdiv_,benchmark/test_addcdiv_.py,,yes,addcdiv_
+main,addcmul,addcmul,benchmark/test_addcmul.py,,yes,addcmul
+main,addcmul_out,addcmul,benchmark/test_addcmul.py,,yes,addcmul_out
+main,addcmul_,addcmul_,benchmark/test_addcmul_.py,,yes,addcmul_
+main,addmm,addmm,benchmark/test_addmm.py,,yes,addmm
+main,addmm_dtype,addmm,benchmark/test_addmm.py,,yes,addmm_dtype
+main,addmm_dtype_out,addmm,benchmark/test_addmm.py,,yes,addmm_dtype_out
+main,addmm_out,addmm,benchmark/test_addmm.py,,yes,addmm_out
+main,addmm_,addmm_,benchmark/test_addmm_.py,,yes,addmm_
+main,addmv,addmv,benchmark/test_addmv.py,,yes,addmv
+main,addmv_out,addmv,benchmark/test_addmv.py,,yes,addmv_out
+main,addr,addr,benchmark/test_addr.py,,yes,addr
+main,affine_grid_generator,affine_grid_generator,benchmark/test_affine_grid_generator.py,,yes,affine_grid_generator
+main,alias,alias,benchmark/test_alias.py,,yes,alias
+main,alias_copy,alias_copy,benchmark/test_alias_copy.py,,yes,alias_copy
+main,alias_copy_out,alias_copy,benchmark/test_alias_copy.py,,yes,alias_copy_out
+main,all,all,benchmark/test_all.py,,yes,all
+main,all_dim,all,benchmark/test_all.py,,yes,all_dim
+main,all_dims,all,benchmark/test_all.py,,yes,all_dims
+main,allclose,allclose,benchmark/test_allclose.py,,yes,allclose
+main,alpha_dropout,alpha_dropout,benchmark/test_alpha_dropout.py,,yes,alpha_dropout
+main,amax,amax,benchmark/test_amax.py,,yes,amax
+main,amin,amin,benchmark/test_amin.py,,yes,amin
+main,amin_,amin,benchmark/test_amin.py,,yes,amin_
+main,aminmax,aminmax,benchmark/test_aminmax.py,,yes,aminmax
+main,amp_foreach_non_finite_check_and_unscale_,amp_foreach_non_finite_check_and_unscale_,benchmark/test_amp_foreach_non_finite_check_and_unscale_.py,,yes,amp_foreach_non_finite_check_and_unscale_
+main,angle,angle,benchmark/test_angle.py,,yes,angle
+main,any,any,benchmark/test_any.py,,yes,any
+main,any_dim,any,benchmark/test_any.py,,yes,any_dim
+main,any_dims,any,benchmark/test_any.py,,yes,any_dims
+main,apply_repetition_penalties,apply_repetition_penalties,benchmark/test_apply_repetition_penalties.py,,yes,apply_repetition_penalties
+main,apply_rotary_pos_emb,apply_rotary_pos_emb,benchmark/test_apply_rotary_pos_emb.py,,yes,apply_rotary_pos_emb
+main,arange,arange,benchmark/test_arange.py,,yes,arange
+main,arange_start,arange,benchmark/test_arange.py,,yes,arange_start
+main,arange_start_step,arange,benchmark/test_arange.py,,yes,arange_start_step
+main,arccos,arccos,benchmark/test_arccos.py,,yes,arccos
+main,arccos_,arccos,benchmark/test_arccos.py,,yes,arccos_
+main,arccosh,arccosh,benchmark/test_arccosh.py,,yes,arccosh
+main,arccosh_out,arccosh,benchmark/test_arccosh.py,,yes,arccosh_out
+main,arccosh_,arccosh_,benchmark/test_arccosh_.py,,yes,arccosh_
+main,arcsin,arcsin,benchmark/test_arcsin.py,,yes,arcsin
+main,arcsin_,arcsin,benchmark/test_arcsin.py,,yes,arcsin_
+main,arcsin_out,arcsin,benchmark/test_arcsin.py,,yes,arcsin_out
+main,arcsinh,arcsinh,benchmark/test_arcsinh.py,,yes,arcsinh
+main,arcsinh_inplace,arcsinh,benchmark/test_arcsinh.py,,yes,arcsinh_
+main,arcsinh_out,arcsinh,benchmark/test_arcsinh.py,,yes,arcsinh_out
+main,arctan2,arctan2,benchmark/test_arctan2.py,,yes,arctan2
+main,arctan2_,arctan2,benchmark/test_arctan2.py,,yes,arctan2_
+main,arctan,arctan_,benchmark/test_arctan_.py,,yes,arctan
+main,arctan_,arctan_,benchmark/test_arctan_.py,,yes,arctan_
+main,arctanh_inplace,arctanh,benchmark/test_arctanh.py,,yes,arctanh_
+main,argmax,argmax,benchmark/test_argmax.py,,yes,argmax
+main,argmin,argmin,benchmark/test_argmin.py,,yes,argmin
+main,argsort,argsort,benchmark/test_argsort.py,,yes,argsort
+main,as_strided_copy,as_strided_copy,benchmark/test_as_strided_copy.py,,yes,as_strided_copy
+main,as_strided_copy_out,as_strided_copy,benchmark/test_as_strided_copy.py,,yes,as_strided_copy_out
+main,asin,asin,benchmark/test_asin.py,,yes,asin
+main,asin_,asin,benchmark/test_asin.py,,yes,asin_
+main,asinh,asinh,benchmark/test_asinh.py,,yes,asinh
+main,asinh_inplace,asinh,benchmark/test_asinh.py,,yes,asinh_
+main,assert_async,assert_async,benchmark/test_assert_async.py,,yes,assert_async
+main,atan,atan,benchmark/test_atan.py,,yes,atan
+main,atan_inplace,atan,benchmark/test_atan.py,,yes,atan_
+main,atan2,atan2,benchmark/test_atan2.py,,yes,atan2
+main,atan2_out,atan2,benchmark/test_atan2.py,,yes,atan2_out
+main,atan2_,atan2_,benchmark/test_atan2_.py,,yes,atan2_
+main,atanh,atanh,benchmark/test_atanh.py,,yes,atanh
+main,atanh_,atanh_,benchmark/test_atanh_.py,,yes,atanh_
+main,avg_pool2d,avg_pool2d,benchmark/test_avg_pool2d.py,,yes,avg_pool2d
+main,avg_pool2d_backward,avg_pool2d,benchmark/test_avg_pool2d.py,,yes,avg_pool2d_backward
+main,perf_avg_pool3d,avg_pool3d_perf,benchmark/test_avg_pool3d_perf.py,,yes,avg_pool3d
+main,perf_avg_pool3d_backward,avg_pool3d_perf,benchmark/test_avg_pool3d_perf.py,,yes,avg_pool3d
+main,baddbmm,baddbmm,benchmark/test_baddbmm.py,,yes,baddbmm
+main,baddbmm_out,baddbmm,benchmark/test_baddbmm.py,,yes,baddbmm_out
+main,baddbmm_,baddbmm_,benchmark/test_baddbmm_.py,,yes,baddbmm_
+main,batch_norm,batch_norm,benchmark/test_batch_norm.py,,yes,batch_norm
+main,batch_norm_backward,batch_norm_backward,benchmark/test_batch_norm_backward.py,,yes,batch_norm_backward
+main,batch_norm_no_update,batch_norm_no_update,benchmark/test_batch_norm_no_update.py,,yes,batch_norm_no_update
+main,beam_search_score,beam_search_score,benchmark/test_beam_search_score.py,,yes,beam_search_score
+main,beam_search_score_,beam_search_score,benchmark/test_beam_search_score.py,,yes,beam_search_score_
+main,bernoulli_inplace,bernoulli,benchmark/test_bernoulli.py,,yes,bernoulli_
+main,bernoulli,bernoulli,benchmark/test_bernoulli.py,,yes,bernoulli
+main,bf16_paged_mqa_logits,bf16_paged_mqa_logits,benchmark/test_bf16_paged_mqa_logits.py,,yes,bf16_paged_mqa_logits
+main,binary_cross_entropy_backward,binary_cross_entropy_backward,benchmark/test_binary_cross_entropy_backward.py,,yes,binary_cross_entropy_backward
+main,binary_cross_entropy_with_logits,binary_cross_entropy_with_logits,benchmark/test_binary_cross_entropy_with_logits.py,,yes,binary_cross_entropy_with_logits
+main,bincount,bincount,benchmark/test_bincount.py,,yes,bincount
+main,bincount_weighted,bincount,benchmark/test_bincount.py,,yes,bincount_weighted
+main,bitwise_and,bitwise_and,benchmark/test_bitwise_and.py,,yes,bitwise_and_tensor
+main,bitwise_and_inplace,bitwise_and,benchmark/test_bitwise_and.py,,yes,bitwise_and_tensor_
+main,bitwise_and_scalar,bitwise_and,benchmark/test_bitwise_and.py,,yes,bitwise_and_scalar
+main,bitwise_and_scalar_,bitwise_and,benchmark/test_bitwise_and.py,,yes,bitwise_and_scalar_
+main,bitwise_and_scalar_tensor,bitwise_and,benchmark/test_bitwise_and.py,,yes,bitwise_and_scalar_tensor
+main,bitwise_left_shift,bitwise_left_shift,benchmark/test_bitwise_left_shift.py,,yes,bitwise_left_shift
+main,bitwise_left_shift_,bitwise_left_shift_,benchmark/test_bitwise_left_shift_.py,,yes,bitwise_left_shift_
+main,bitwise_not,bitwise_not,benchmark/test_bitwise_not.py,,yes,bitwise_not
+main,bitwise_not_inplace,bitwise_not,benchmark/test_bitwise_not.py,,yes,bitwise_not_
+main,bitwise_or_tensor,bitwise_or,benchmark/test_bitwise_or.py,,yes,bitwise_or_tensor
+main,bitwise_or_inplace,bitwise_or,benchmark/test_bitwise_or.py,,yes,bitwise_or_tensor_
+main,bitwise_or_scalar,bitwise_or,benchmark/test_bitwise_or.py,,yes,bitwise_or_scalar
+main,bitwise_or_scalar_,bitwise_or,benchmark/test_bitwise_or.py,,yes,bitwise_or_scalar_
+main,bitwise_or_scalar_tensor,bitwise_or,benchmark/test_bitwise_or.py,,yes,bitwise_or_scalar_tensor
+main,bitwise_right_shift,bitwise_right_shift,benchmark/test_bitwise_right_shift.py,,yes,bitwise_right_shift
+main,bitwise_right_shift_,bitwise_right_shift_,benchmark/test_bitwise_right_shift_.py,,yes,bitwise_right_shift_
+main,bitwise_xor_tensor,bitwise_xor,benchmark/test_bitwise_xor.py,,yes,bitwise_xor_tensor
+main,bitwise_xor_inplace,bitwise_xor,benchmark/test_bitwise_xor.py,,yes,bitwise_xor_tensor_
+main,bitwise_xor_scalar,bitwise_xor,benchmark/test_bitwise_xor.py,,yes,bitwise_xor_scalar
+main,bitwise_xor_scalar_,bitwise_xor,benchmark/test_bitwise_xor.py,,yes,bitwise_xor_scalar_
+main,bitwise_xor_scalar_tensor,bitwise_xor,benchmark/test_bitwise_xor.py,,yes,bitwise_xor_scalar_tensor
+main,blas_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,
+main,perf_w8a8_block_fp8_matmul,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,w8a8_block_fp8_matmul
+main,perf_w8a8_block_fp8_matmul_deepgemm,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,w8a8_block_fp8_matmul_deepgemm
+main,perf_sparse_attention,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,sparse_attention
+main,perf_mul,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,mul
+main,mv_and_outer_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,
+main,addmv_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,
+main,vdot_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,vdot
+main,addr_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,addr
+main,perf_router_gemm,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,router_gemm
+main,block_diag,block_diag,benchmark/test_block_diag.py,,yes,block_diag
+main,bmm,bmm,benchmark/test_bmm.py,,yes,bmm
+main,bmm_out,bmm,benchmark/test_bmm.py,,yes,bmm_out
+main,broadcast_tensors,broadcast_tensors,benchmark/test_broadcast_tensors.py,,yes,broadcast_tensors
+main,broadcast_to,broadcast_to,benchmark/test_broadcast_to.py,,yes,broadcast_to
+main,bucketize_perf,bucketize,benchmark/test_bucketize.py,,yes,bucketize
+main,cat,cat,benchmark/test_cat.py,,yes,cat
+main,cat_out,cat,benchmark/test_cat.py,,yes,cat_out
+main,cauchy_inplace,cauchy,benchmark/test_cauchy.py,,yes,cauchy_
+main,cauchy_out,cauchy,benchmark/test_cauchy.py,,yes,cauchy
+main,cdist_backward,cdist_backward,benchmark/test_cdist_backward.py,,yes,cdist_backward
+main,cdist_forward,cdist_forward,benchmark/test_cdist_forward.py,,yes,cdist_forward
+main,ceil,ceil,benchmark/test_ceil.py,,yes,ceil
+main,ceil_inplace,ceil,benchmark/test_ceil.py,,yes,ceil_
+main,ceil_out,ceil,benchmark/test_ceil.py,,yes,ceil_out
+main,celu,celu,benchmark/test_celu.py,,yes,celu
+main,celu_inplace,celu,benchmark/test_celu.py,,yes,celu_
+main,channel_shuffle,channel_shuffle,benchmark/test_channel_shuffle.py,,yes,channel_shuffle
+main,cholesky_inverse,cholesky_inverse,benchmark/test_cholesky_inverse.py,,yes,cholesky_inverse
+main,cholesky_solve,cholesky_solve,benchmark/test_cholesky_solve.py,,yes,cholesky_solve
+main,cholesky_solve_out,cholesky_solve,benchmark/test_cholesky_solve.py,,yes,cholesky_solve_out
+main,chunk,chunk,benchmark/test_chunk.py,,yes,chunk
+main,chunk_cat,chunk_cat,benchmark/test_chunk_cat.py,,yes,chunk_cat
+main,clamp,clamp,benchmark/test_clamp.py,,yes,clamp
+main,clamp_inplace,clamp,benchmark/test_clamp.py,,yes,clamp_
+main,clamp_tensor,clamp,benchmark/test_clamp.py,,yes,clamp_tensor
+main,clamp_tensor_inplace,clamp,benchmark/test_clamp.py,,yes,clamp_tensor_
+main,clamp_max,clamp_max,benchmark/test_clamp_max.py,,yes,clamp_max
+main,clamp_max_inplace,clamp_max,benchmark/test_clamp_max.py,,yes,clamp_max_
+main,clamp_min,clamp_min,benchmark/test_clamp_min.py,,yes,clamp_min
+main,clamp_min_inplace,clamp_min,benchmark/test_clamp_min.py,,yes,clamp_min_
+main,clip,clip,benchmark/test_clip.py,,yes,clip
+main,clip_inplace,clip,benchmark/test_clip.py,,yes,clip_
+main,col2im,col2im,benchmark/test_col2im.py,,yes,col2im
+main,concat_and_cache_mla,concat_and_cache_mla,benchmark/test_concat_and_cache_mla.py,,yes,concat_and_cache_mla
+main,concatenate,concatenate,benchmark/test_concatenate.py,,yes,concatenate
+main,conj,conj,benchmark/test_conj.py,,yes,conj
+main,conj_physical,conj_physical,benchmark/test_conj_physical.py,,yes,conj_physical
+main,constant_pad_nd,constant_pad_nd,benchmark/test_constant_pad_nd.py,,yes,constant_pad_nd
+main,contiguous,contiguous,benchmark/test_contiguous.py,,yes,contiguous
+main,conv1d,conv1d,benchmark/test_conv1d.py,,yes,conv1d
+main,conv1d_padding,conv1d,benchmark/test_conv1d.py,,yes,conv1d_padding
+main,conv2d,conv2d,benchmark/test_conv2d.py,,yes,conv2d
+main,conv2d_padding,conv2d,benchmark/test_conv2d.py,,yes,conv2d_padding
+main,conv3d,conv3d,benchmark/test_conv3d.py,,yes,conv3d
+main,conv3d_padding,conv3d,benchmark/test_conv3d.py,,yes,conv3d_padding
+main,conv_depthwise2d,conv_depthwise2d,benchmark/test_conv_depthwise2d.py,,yes,conv_depthwise2d
+main,conv_transpose1d,conv_transpose1d,benchmark/test_conv_transpose1d.py,,yes,conv_transpose1d
+main,perf_conv_transpose2d,conv_transpose2d,benchmark/test_conv_transpose2d.py,,yes,conv_transpose2d
+main,convert_weight_to_int4pack,convert_weight_to_int4pack,benchmark/test_convert_weight_to_int4pack.py,,yes,convert_weight_to_int4pack
+main,copy_inplace,copy,benchmark/test_copy.py,,yes,copy_
+main,copy_functional,copy,benchmark/test_copy.py,,yes,copy
+main,copysign,copysign,benchmark/test_copysign.py,,yes,copysign
+main,copysign_out,copysign,benchmark/test_copysign.py,,yes,copysign_out
+main,copysign_,copysign_,benchmark/test_copysign_.py,,yes,copysign_
+main,cos,cos,benchmark/test_cos.py,,yes,cos
+main,cos_inplace,cos,benchmark/test_cos.py,,yes,cos_
+main,cosh,cosh,benchmark/test_cosh.py,,yes,cosh
+main,cosh_inplace,cosh,benchmark/test_cosh.py,,yes,cosh_
+main,cosh_out,cosh,benchmark/test_cosh.py,,yes,cosh_out
+main,count_nonzero,count_nonzero,benchmark/test_count_nonzero.py,,yes,count_nonzero
+main,cp_gather_indexer_k_quant_cache_benchmark,cp_gather_indexer_k_quant_cache,benchmark/test_cp_gather_indexer_k_quant_cache.py,,yes,
+main,cross_entropy_loss,cross_entropy_loss,benchmark/test_cross_entropy_loss.py,,yes,cross_entropy_loss
+main,ctc_loss,ctc_loss,benchmark/test_ctc_loss.py,,yes,ctc_loss
+main,ctc_loss_backward,ctc_loss,benchmark/test_ctc_loss.py,,yes,ctc_loss
+main,perf_cudnn_attention_forward,cudnn_attention_forward,benchmark/test_cudnn_attention_forward.py,,yes,cudnn_attention_forward
+main,cudnn_batch_norm_backward,cudnn_batch_norm_backward,benchmark/test_cudnn_batch_norm_backward.py,,yes,cudnn_batch_norm_backward
+main,cudnn_convolution,cudnn_convolution,benchmark/test_cudnn_convolution.py,,yes,cudnn_convolution
+main,cummax,cummax,benchmark/test_cummax.py,,yes,cummax
+main,cummin,cummin,benchmark/test_cummin.py,,yes,cummin
+main,cumprod_perf,cumprod,benchmark/test_cumprod.py,,yes,cumprod
+main,cumprod_inplace_perf,cumprod,benchmark/test_cumprod.py,,yes,cumprod_
+main,cumsum,cumsum,benchmark/test_cumsum.py,,yes,cumsum
+main,cumsum_out,cumsum,benchmark/test_cumsum.py,,yes,cumsum_out
+main,cutlass_scaled_mm_benchmark,cutlass_scaled_mm,benchmark/test_cutlass_scaled_mm.py,,yes,
+main,combine_topk_swa_indices_benchmark,deepseek_v4_attention_combine_topk_swa_indices,benchmark/test_deepseek_v4_attention_combine_topk_swa_indices.py,,yes,
+main,compute_global_topk_indices_and_lens_benchmark,deepseek_v4_attention_compute_global_topk_indices_and_lens,benchmark/test_deepseek_v4_attention_compute_global_topk_indices_and_lens.py,,yes,
+main,dequantize_and_gather_k_cache_benchmark,deepseek_v4_attention_dequantize_and_gather_k_cache,benchmark/test_deepseek_v4_attention_dequantize_and_gather_k_cache.py,,yes,
+main,fused_q_kv_rmsnorm_benchmark,deepseek_v4_attention_fused_q_kv_rmsnorm,benchmark/test_deepseek_v4_attention_fused_q_kv_rmsnorm.py,,yes,
+main,deg2rad,deg2rad,benchmark/test_deg2rad.py,,yes,deg2rad
+main,deg2rad_,deg2rad,benchmark/test_deg2rad.py,,yes,deg2rad_
+main,deg2rad_out,deg2rad,benchmark/test_deg2rad.py,,yes,deg2rad_out
+main,dequantize,dequantize,benchmark/test_dequantize.py,,yes,dequantize
+main,dgeglu,dgeglu,benchmark/test_dgeglu.py,,yes,dgeglu
+main,diag,diag,benchmark/test_diag.py,,yes,diag
+main,diag_embed,diag_embed,benchmark/test_diag_embed.py,,yes,diag_embed
+main,diagonal_backward,diagonal,benchmark/test_diagonal.py,,yes,diagonal_backward
+main,diagonal_copy,diagonal_copy,benchmark/test_diagonal_copy.py,,yes,diagonal_copy
+main,diagonal_scatter,diagonal_scatter,benchmark/test_diagonal_scatter.py,,yes,diagonal_scatter
+main,diff,diff,benchmark/test_diff.py,,yes,diff
+main,diff_n2,diff,benchmark/test_diff.py,,yes,diff
+main,digamma,digamma,benchmark/test_digamma.py,,yes,digamma
+main,digamma_inplace,digamma,benchmark/test_digamma.py,,yes,digamma_
+main,div,div,benchmark/test_div.py,,yes,div_tensor
+main,div_inplace,div,benchmark/test_div.py,,yes,div_tensor_
+main,div_tensor_mode,div,benchmark/test_div.py,,yes,div_tensor_mode
+main,div_tensor_mode_inplace,div,benchmark/test_div.py,,yes,div_tensor_mode_
+main,div_scalar_mode,div,benchmark/test_div.py,,yes,div_scalar_mode
+main,div_scalar_mode_inplace,div,benchmark/test_div.py,,yes,div_scalar_mode_
+main,div_scalar_inplace,div,benchmark/test_div.py,,yes,div_scalar_
+main,div_out,div,benchmark/test_div.py,,yes,div_out
+main,dot,dot,benchmark/test_dot.py,,yes,dot
+main,dreglu,dreglu,benchmark/test_dreglu.py,,yes,dreglu
+main,dropout,dropout,benchmark/test_dropout.py,,yes,dropout
+main,dropout_backward,dropout,benchmark/test_dropout.py,,yes,dropout_backward
+main,dswiglu,dswiglu,benchmark/test_dswiglu.py,,yes,dswiglu
+main,dunder_ior_inplace,dunder_ior,benchmark/test_dunder_ior.py,,yes,dunder_ior_tensor
+main,dunder_ior_scalar_inplace,dunder_ior,benchmark/test_dunder_ior.py,,yes,dunder_ior_scalar
+main,dunder_or_tensor,dunder_or,benchmark/test_dunder_or.py,,yes,dunder_or_tensor
+main,dunder_or_scalar,dunder_or,benchmark/test_dunder_or.py,,yes,dunder_or_scalar
+main,dyn_quant_pack_4bit_weight,dyn_quant_pack_4bit_weight,benchmark/test_dyn_quant_pack_4bit_weight.py,,yes,dyn_quant_pack_4bit_weight
+main,einsum_matmul,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_bmm,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_dot,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_outer,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_trace,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_diagonal,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_transpose,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_sum_all,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_sum_dim,einsum,benchmark/test_einsum.py,,yes,einsum
+main,einsum_ellipsis,einsum,benchmark/test_einsum.py,,yes,einsum
+main,elu,elu,benchmark/test_elu.py,,yes,elu
+main,elu_inplace,elu,benchmark/test_elu.py,,yes,elu_
+main,elu_backward,elu,benchmark/test_elu.py,,yes,elu_backward
+main,embedding,embedding,benchmark/test_embedding.py,,yes,embedding
+main,embedding_backward,embedding,benchmark/test_embedding.py,,yes,embedding_backward
+main,embedding_bag_dense_backward,embedding_bag_dense_backward,benchmark/test_embedding_bag_dense_backward.py,,yes,embedding_bag_dense_backward
+main,embedding_bag_per_sample_weights_backward,embedding_bag_per_sample_weights_backward,benchmark/test_embedding_bag_per_sample_weights_backward.py,,yes,_embedding_bag_per_sample_weights_backward
+main,embedding_dense_backward,embedding_dense_backward,benchmark/test_embedding_dense_backward.py,,yes,embedding_dense_backward
+main,empty,empty,benchmark/test_empty.py,,yes,empty
+main,eq,eq,benchmark/test_eq.py,,yes,eq
+main,eq_scalar,eq,benchmark/test_eq.py,,yes,eq_scalar
+main,eq_,eq_,benchmark/test_eq_.py,,yes,eq_
+main,eq_scalar_,eq_,benchmark/test_eq_.py,,yes,eq_scalar_
+main,equal,equal,benchmark/test_equal.py,,yes,equal
+main,special_erf,erf,benchmark/test_erf.py,,yes,special_erf
+main,erf,erf,benchmark/test_erf.py,,yes,erf
+main,erf_inplace,erf,benchmark/test_erf.py,,yes,erf_
+main,erfinv,erfinv_,benchmark/test_erfinv_.py,,yes,erfinv
+main,erfinv_,erfinv_,benchmark/test_erfinv_.py,,yes,erfinv_
+main,euclidean_dist,euclidean_dist,benchmark/test_euclidean_dist.py,,yes,euclidean_dist
+main,exp,exp,benchmark/test_exp.py,,yes,exp
+main,exp_inplace,exp,benchmark/test_exp.py,,yes,exp_
+main,exp_out,exp,benchmark/test_exp.py,,yes,exp_out
+main,special_exp2,exp2,benchmark/test_exp2.py,,yes,special_exp2
+main,exp2,exp2,benchmark/test_exp2.py,,yes,exp2
+main,exp2_inplace,exp2,benchmark/test_exp2.py,,yes,exp2_
+main,expand,expand,benchmark/test_expand.py,,yes,expand
+main,expand_,expand,benchmark/test_expand.py,,yes,expand_
+main,expand_as,expand_as,benchmark/test_expand_as.py,,yes,expand_as
+main,expm1,expm1,benchmark/test_expm1.py,,yes,expm1
+main,expm1_inplace,expm1,benchmark/test_expm1.py,,yes,expm1_
+main,expm1_out,expm1,benchmark/test_expm1.py,,yes,expm1_out
+main,exponential_inplace,exponential,benchmark/test_exponential.py,,yes,exponential_
+main,eye,eye,benchmark/test_eye.py,,yes,eye
+main,eye_m,eye,benchmark/test_eye.py,,yes,eye_m
+main,feature_dropout,feature_dropout,benchmark/test_feature_dropout.py,,yes,feature_dropout
+main,feature_dropout_,feature_dropout,benchmark/test_feature_dropout.py,,yes,feature_dropout_
+main,perf_fft,fft,benchmark/test_fft.py,,yes,fft
+main,fill_scalar,fill,benchmark/test_fill.py,,yes,fill_scalar
+main,fill_tensor,fill,benchmark/test_fill.py,,yes,fill_tensor
+main,fill_tensor_inplace,fill,benchmark/test_fill.py,,yes,fill_tensor_
+main,fill_tensor_out,fill,benchmark/test_fill.py,,yes,fill_tensor_out
+main,fill_scalar_inplace,fill,benchmark/test_fill.py,,yes,fill_scalar_
+main,fill_scalar_out,fill,benchmark/test_fill.py,,yes,fill_scalar_out
+main,fix,fix,benchmark/test_fix.py,,yes,fix
+main,fix_,fix_,benchmark/test_fix_.py,,yes,fix_
+main,perf_flash_attention_backward,flash_attention_backward,benchmark/test_flash_attention_backward.py,,yes,flash_attention_backward
+main,perf_scaled_dot_product_flash_attention_backward,flash_attention_backward,benchmark/test_flash_attention_backward.py,,yes,scaled_dot_product_flash_attention_backward
+main,perf_scaled_dot_product_cudnn_attention_backward,flash_attention_backward,benchmark/test_flash_attention_backward.py,,yes,scaled_dot_product_cudnn_attention_backward
+main,perf_efficient_attention_backward,flash_attention_backward,benchmark/test_flash_attention_backward.py,,yes,efficient_attention_backward
+main,perf_scaled_dot_product_efficient_attention_backward,flash_attention_backward,benchmark/test_flash_attention_backward.py,,yes,scaled_dot_product_efficient_attention_backward
+main,_flash_attention_forward,flash_attention_forward,benchmark/test_flash_attention_forward.py,,yes,_flash_attention_forward
+main,flash_attention_forward,flash_attention_forward,benchmark/test_flash_attention_forward.py,,yes,flash_attention_forward
+main,flash_attn_varlen_func,flash_attn_varlen_func,benchmark/test_flash_attn_varlen_func.py,,yes,flash_attn_varlen_func
+main,flash_attn_varlen_func_noncontig,flash_attn_varlen_func,benchmark/test_flash_attn_varlen_func.py,,yes,flash_attn_varlen_func_noncontig
+main,flash_attn_varlen_opt_func,flash_attn_varlen_opt_init_func,benchmark/test_flash_attn_varlen_opt_init_func.py,,yes,flash_attn_varlen_func
+main,flash_mla,flash_mla,benchmark/test_flash_mla.py,,yes,flash_mla
+main,flash_mla_sparse_fwd,flash_mla_sparse_fwd,benchmark/test_flash_mla_sparse_fwd.py,,yes,
+main,flash_mla_with_kvcache,flash_mla_with_kvcache,benchmark/test_flash_mla_with_kvcache.py,,yes,
+main,flatten,flatten,benchmark/test_flatten.py,,yes,flatten.using_ints
+main,flip,flip,benchmark/test_flip.py,,yes,flip
+main,float_power_,float_power_,benchmark/test_float_power_.py,,yes,float_power_
+main,floor,floor,benchmark/test_floor.py,,yes,floor
+main,floor_inplace,floor,benchmark/test_floor.py,,yes,floor_
+main,floor_out,floor,benchmark/test_floor.py,,yes,floor_out
+main,floor_divide,floor_divide,benchmark/test_floor_divide.py,,yes,floor_divide
+main,floor_divide_inplace,floor_divide,benchmark/test_floor_divide.py,,yes,floor_divide_
+main,floor_divide_scalar,floor_divide,benchmark/test_floor_divide.py,,yes,floor_divide_scalar
+main,floor_divide_scalar_,floor_divide,benchmark/test_floor_divide.py,,yes,floor_divide_scalar_
+main,floor_divide_tensor,floor_divide,benchmark/test_floor_divide.py,,yes,floor_divide_tensor
+main,floor_divide_tensor_inplace,floor_divide,benchmark/test_floor_divide.py,,yes,floor_divide_tensor_
+main,fmax,fmax,benchmark/test_fmax.py,,yes,fmax
+main,fmax_out,fmax,benchmark/test_fmax.py,,yes,fmax_out
+main,fmin,fmin,benchmark/test_fmin.py,,yes,fmin
+main,fmin_out,fmin,benchmark/test_fmin.py,,yes,fmin_out
+main,fmod_tensor,fmod,benchmark/test_fmod.py,,yes,fmod_tensor
+main,fmod_scalar,fmod,benchmark/test_fmod.py,,yes,fmod_scalar
+main,fmod_tensor_,fmod,benchmark/test_fmod.py,,yes,fmod_tensor_
+main,fmod_scalar_,fmod,benchmark/test_fmod.py,,yes,fmod_scalar_
+main,fmod_,fmod_,benchmark/test_fmod_.py,,yes,fmod_
+main,perf_fp8_einsum_gems_vs_deepgemm,fp8_einsum,benchmark/test_fp8_einsum.py,,yes,fp8_einsum
+main,fp8_fp4_mega_moe_vllm,fp8_fp4_mega_moe,benchmark/test_fp8_fp4_mega_moe.py,,yes,fp8_fp4_mega_moe
+main,fp8_fp4_mega_moe_torch_ref,fp8_fp4_mega_moe,benchmark/test_fp8_fp4_mega_moe.py,,yes,fp8_fp4_mega_moe_torch_ref
+main,fp8_fp4_mqa_logits,fp8_fp4_mqa_logits,benchmark/test_fp8_fp4_mqa_logits.py,,yes,fp8_fp4_mqa_logits
+main,fp8_fp4_paged_mqa_logits,fp8_fp4_paged_mqa_logits,benchmark/test_fp8_fp4_paged_mqa_logits.py,,yes,fp8_fp4_paged_mqa_logits
+main,fp8_mqa_logits,fp8_mqa_logits,benchmark/test_fp8_mqa_logits.py,,yes,fp8_mqa_logits
+main,perf_fp8_paged_mqa_logits_gems_vs_deepgemm,fp8_paged_mqa_logits_perf,benchmark/test_fp8_paged_mqa_logits_perf.py,,yes,
+main,frac,frac,benchmark/test_frac.py,,yes,frac
+main,frac_,frac_,benchmark/test_frac_.py,,yes,frac_
+main,fractional_max_pool2d,fractional_max_pool2d,benchmark/test_fractional_max_pool2d.py,,yes,fractional_max_pool2d
+main,fractional_max_pool2d_backward,fractional_max_pool2d,benchmark/test_fractional_max_pool2d.py,,yes,fractional_max_pool2d_backward
+main,full,full,benchmark/test_full.py,,yes,full
+main,full_like,full_like,benchmark/test_full_like.py,,yes,full_like
+main,functional_assert_async,functional_assert_async,benchmark/test_functional_assert_async.py,,yes,functional_assert_async
+main,functional_sym_constrain_range,functional_sym_constrain_range,benchmark/test_functional_sym_constrain_range.py,,yes,functional_sym_constrain_range
+main,functional_sym_constrain_range_for_size,functional_sym_constrain_range_for_size,benchmark/test_functional_sym_constrain_range_for_size.py,,yes,functional_sym_constrain_range_for_size
+main,fused_adam,fused_adam,benchmark/test_fused_adam.py,,yes,fused_adam
+main,fused_adam_,fused_adam,benchmark/test_fused_adam.py,,yes,fused_adam_
+main,fused_add_rms_norm,fused_add_rms_norm,benchmark/test_fused_add_rms_norm.py,,yes,fused_add_rms_norm
+main,fused_deepseek_v4_qnorm_rope_kv_rope_insert,fused_deepseek_v4_qnorm_rope_kv_rope_insert,benchmark/test_fused_deepseek_v4_qnorm_rope_kv_rope_insert.py,,yes,fused_deepseek_v4_qnorm_rope_kv_rope_insert
+main,fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert,fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert,benchmark/test_fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert.py,,yes,
+main,fused_indexer_q_rope_quant,fused_indexer_q_rope_quant,benchmark/test_fused_indexer_q_rope_quant.py,,yes,fused_indexer_q_rope_quant
+main,fused_inv_rope_fp8_quant,fused_inv_rope_fp8_quant,benchmark/test_fused_inv_rope_fp8_quant.py,,yes,fused_inv_rope_fp8_quant
+main,fused_marlin_moe,fused_marlin_moe,benchmark/test_fused_marlin_moe.py,,yes,fused_marlin_moe
+main,fused_marlin_moe_int8,fused_marlin_moe,benchmark/test_fused_marlin_moe.py,,yes,fused_marlin_moe_int8
+main,fused_marlin_moe_mxfp4,fused_marlin_moe,benchmark/test_fused_marlin_moe.py,,yes,fused_marlin_moe_mxfp4
+main,fused_moe_impl_gems_vs_vllm,fused_moe,benchmark/test_fused_moe.py,,yes,fused_experts_impl
+main,dispatch_fused_moe_kernel,fused_moe,benchmark/test_fused_moe.py,,yes,dispatch_fused_moe_kernel
+main,fused_moe_fp8,fused_moe_fp8,benchmark/test_fused_moe_fp8.py,,yes,fused_experts_impl
+main,fused_moe_fp8_blockwise,fused_moe_fp8_blockwise,benchmark/test_fused_moe_fp8_blockwise.py,,yes,fused_experts_impl
+main,fused_moe_int4_w4a16,fused_moe_int4_w4a16,benchmark/test_fused_moe_int4_w4a16.py,,yes,fused_experts_impl
+main,fused_experts_impl_int8,fused_moe_int8,benchmark/test_fused_moe_int8.py,,yes,fused_experts_impl
+main,fused_experts_impl_int8_w8a16,fused_moe_int8_w8a16,benchmark/test_fused_moe_int8_w8a16.py,,yes,fused_moe_int8_w8a16_gems_vs_bf16_deq
+main,fused_moe_w8a16_mxq,fused_moe_w8a16,benchmark/test_fused_moe_w8a16.py,,yes,fused_moe
+main,fused_moe_w8a16_mxq_gems_vs_vllm,fused_moe_w8a16,benchmark/test_fused_moe_w8a16.py,,yes,fused_moe
+main,fused_rms_norm,fused_rms_norm,benchmark/test_fused_rms_norm.py,,yes,fused_rms_norm
+main,gather,gather,benchmark/test_gather.py,,yes,gather
+main,gather_backward,gather,benchmark/test_gather.py,,yes,gather_backward
+main,gather_block_quantized,gather_block_quantized,benchmark/test_gather_block_quantized.py,,yes,gather_block_quantized
+main,gcd,gcd,benchmark/test_gcd.py,,yes,gcd
+main,gcd_out,gcd,benchmark/test_gcd.py,,yes,gcd_out
+main,gcd_,gcd_,benchmark/test_gcd_.py,,yes,gcd_
+main,ge,ge,benchmark/test_ge.py,,yes,ge
+main,ge_scalar,ge,benchmark/test_ge.py,,yes,ge_scalar
+main,geglu,geglu,benchmark/test_geglu.py,,yes,geglu
+main,gelu,gelu,benchmark/test_gelu.py,,yes,gelu
+main,gelu_inplace,gelu,benchmark/test_gelu.py,,yes,gelu_
+main,gelu_backward,gelu,benchmark/test_gelu.py,,yes,gelu_backward
+main,gelu_and_mul,gelu_and_mul,benchmark/test_gelu_and_mul.py,,yes,gelu_and_mul
+main,geometric_inplace,geometric,benchmark/test_geometric.py,,yes,geometric_
+main,geometric,geometric,benchmark/test_geometric.py,,yes,geometric
+main,get_paged_mqa_logits_metadata,get_paged_mql_logits_metadata,benchmark/test_get_paged_mql_logits_metadata.py,,yes,get_paged_mqa_logits_metadata
+main,get_scheduler_metadata,get_scheduler_metadata,benchmark/test_get_scheduler_metadata.py,,yes,get_scheduler_metadata
+main,glu,glu,benchmark/test_glu.py,,yes,glu
+main,glu_backward,glu,benchmark/test_glu.py,,yes,glu_backward
+main,greater,greater,benchmark/test_greater.py,,yes,greater
+main,greater_out,greater,benchmark/test_greater.py,,yes,greater_out
+main,greater_scalar,greater,benchmark/test_greater.py,,yes,greater_scalar
+main,greater_scalar_out,greater,benchmark/test_greater.py,,yes,greater_scalar_out
+main,greater_equal,greater_equal,benchmark/test_greater_equal.py,,yes,greater_equal
+main,greater_equal_,greater_equal,benchmark/test_greater_equal.py,,yes,greater_equal_
+main,greater_equal_scalar,greater_equal,benchmark/test_greater_equal.py,,yes,greater_equal_scalar
+main,grid_sample,grid_sample,benchmark/test_grid_sample.py,,yes,grid_sample
+main,group_norm,group_norm,benchmark/test_group_norm.py,,yes,group_norm
+main,grouped_mm,grouped_mm,benchmark/test_grouped_mm.py,,yes,grouped_mm
+main,grouped_topk_no_renorm,grouped_topk,benchmark/test_grouped_topk.py,,yes,grouped_topk
+main,grouped_topk_score_0,grouped_topk,benchmark/test_grouped_topk.py,,yes,grouped_topk
+main,grouped_topk_score_1,grouped_topk,benchmark/test_grouped_topk.py,,yes,grouped_topk
+main,gt,gt,benchmark/test_gt.py,,yes,gt
+main,gt_scalar,gt,benchmark/test_gt.py,,yes,gt_scalar
+main,gt_scalar_,gt_scalar_,benchmark/test_gt_scalar_.py,,yes,gt_scalar_
+main,gt_tensor_,gt_tensor_,benchmark/test_gt_tensor_.py,,yes,gt_tensor_
+main,hadamard_transform,hadamard_transform,benchmark/test_hadamard_transform.py,,yes,hadamard_transform
+main,hadamard_transform_mn,hadamard_transform,benchmark/test_hadamard_transform.py,,yes,hadamard_transform_mn
+main,hardshrink,hardshrink,benchmark/test_hardshrink.py,,yes,hardshrink
+main,hardshrink_out,hardshrink,benchmark/test_hardshrink.py,,yes,hardshrink_out
+main,hardsigmoid_inplace,hardsigmoid,benchmark/test_hardsigmoid.py,,yes,hardsigmoid_
+main,hardsigmoid,hardsigmoid,benchmark/test_hardsigmoid.py,,yes,hardsigmoid
+main,hardsigmoid_out,hardsigmoid,benchmark/test_hardsigmoid.py,,yes,hardsigmoid_out
+main,hardswish_inplace,hardswish,benchmark/test_hardswish.py,,yes,hardswish_
+main,hardtanh_inplace,hardtanh_,benchmark/test_hardtanh_.py,,yes,hardtanh_
+main,has_compatible_shallow_copy_type,has_compatible_shallow_copy_type,benchmark/test_has_compatible_shallow_copy_type.py,,yes,has_compatible_shallow_copy_type
+main,heaviside_,heaviside_,benchmark/test_heaviside_.py,,yes,heaviside_
+main,histc,histc,benchmark/test_histc.py,,yes,histc
+main,hstack,hstack,benchmark/test_hstack.py,,yes,hstack
+main,huber_loss,huber_loss,benchmark/test_huber_loss.py,,yes,huber_loss
+main,hypot,hypot,benchmark/test_hypot.py,,yes,hypot
+main,hypot_out,hypot,benchmark/test_hypot.py,,yes,hypot_out
+main,hypot_,hypot_,benchmark/test_hypot_.py,,yes,hypot_
+main,i0,i0,benchmark/test_i0.py,,yes,i0
+main,i0_inplace,i0,benchmark/test_i0.py,,yes,i0_
+main,i0_out,i0,benchmark/test_i0.py,,yes,i0_out
+main,igamma_,igamma_,benchmark/test_igamma_.py,,yes,igamma_
+main,igammac,igammac,benchmark/test_igammac.py,,yes,igammac
+main,igammac_out,igammac,benchmark/test_igammac.py,,yes,igammac_out
+main,igammac_,igammac_,benchmark/test_igammac_.py,,yes,igammac_
+main,ilshift,ilshift,benchmark/test_ilshift.py,,yes,ilshift
+main,im2col,im2col,benchmark/test_im2col.py,,yes,im2col
+main,index_acc_perf,index,benchmark/test_index.py,,yes,index
+main,index_add,index_add,benchmark/test_index_add.py,,yes,index_add
+main,index_add_,index_add,benchmark/test_index_add.py,,yes,index_add_
+main,index_copy,index_copy,benchmark/test_index_copy.py,,yes,index_copy
+main,index_copy_,index_copy,benchmark/test_index_copy.py,,yes,index_copy_
+main,index_put_acc_false,index_put,benchmark/test_index_put.py,,yes,index_put
+main,index_put__acc_false,index_put,benchmark/test_index_put.py,,yes,index_put_
+main,index_put_acc_true,index_put,benchmark/test_index_put.py,,yes,index_put
+main,index_put__acc_true,index_put,benchmark/test_index_put.py,,yes,index_put_
+main,index_put_impl_acc_false,index_put_impl,benchmark/test_index_put_impl.py,,yes,index_put_impl
+main,index_put_impl_acc_true,index_put_impl,benchmark/test_index_put_impl.py,,yes,index_put_impl
+main,index_reduce_prod,index_reduce,benchmark/test_index_reduce.py,,yes,
+main,index_reduce_mean,index_reduce,benchmark/test_index_reduce.py,,yes,
+main,index_reduce_amax,index_reduce,benchmark/test_index_reduce.py,,yes,
+main,index_reduce_amin,index_reduce,benchmark/test_index_reduce.py,,yes,
+main,index_select,index_select,benchmark/test_index_select.py,,yes,index_select
+main,index_select_backward,index_select_backward,benchmark/test_index_select_backward.py,,yes,index_select_backward
+main,indexer_k_quant_and_cache_benchmark,indexer_k_quant_and_cache,benchmark/test_indexer_k_quant_and_cache.py,,yes,
+main,inplace_fused_experts_gems_vs_vllm,inplace_fused_experts,benchmark/test_inplace_fused_experts.py,,yes,inplace_fused_experts
+main,instance_norm,instance_norm,benchmark/test_instance_norm.py,,yes,instance_norm
+main,irshift__,irshift,benchmark/test_irshift.py,,yes,irshift__
+main,is_all_true,is_all_true,benchmark/test_is_all_true.py,,yes,is_all_true
+main,is_nonzero,is_nonzero,benchmark/test_is_nonzero.py,,yes,is_nonzero
+main,isclose,isclose,benchmark/test_isclose.py,,yes,isclose
+main,isfinite,isfinite,benchmark/test_isfinite.py,,yes,isfinite
+main,isin,isin,benchmark/test_isin.py,,yes,isin
+main,isin_scalar_tensor,isin,benchmark/test_isin.py,,yes,isin_scalar_tensor
+main,isin_tensor_scalar,isin,benchmark/test_isin.py,,yes,isin_tensor_scalar
+main,isinf,isinf,benchmark/test_isinf.py,,yes,isinf
+main,isnan,isnan,benchmark/test_isnan.py,,yes,isnan
+main,isneginf,isneginf,benchmark/test_isneginf.py,,yes,isneginf
+main,isneginf_out,isneginf,benchmark/test_isneginf.py,yes,yes,isneginf_out
+main,isposinf,isposinf,benchmark/test_isposinf.py,,yes,isposinf
+main,jagged_to_padded_dense_forward,jagged_to_padded_dense_forward,benchmark/test_jagged_to_padded_dense_forward.py,,yes,jagged_to_padded_dense_forward
+main,kron,kron,benchmark/test_kron.py,,yes,kron
+main,kthvalue,kthvalue,benchmark/test_kthvalue.py,,yes,kthvalue
+main,native_layer_norm,layer_norm,benchmark/test_layer_norm.py,,yes,native_layer_norm
+main,layer_norm,layer_norm,benchmark/test_layer_norm.py,,yes,layer_norm
+main,layer_norm_backward,layer_norm,benchmark/test_layer_norm.py,,yes,layer_norm_backward
+main,lcm,lcm,benchmark/test_lcm.py,,yes,lcm
+main,lcm_,lcm_,benchmark/test_lcm_.py,,yes,lcm_
+main,le,le,benchmark/test_le.py,,yes,le
+main,le_scalar,le,benchmark/test_le.py,,yes,le_scalar
+main,le_,le_,benchmark/test_le_.py,,yes,le_
+main,le_scalar_,le_,benchmark/test_le_.py,,yes,le_scalar_
+main,leaky_relu,leaky_relu,benchmark/test_leaky_relu.py,,yes,leaky_relu
+main,leaky_relu_inplace,leaky_relu,benchmark/test_leaky_relu.py,,yes,leaky_relu_
+main,leaky_relu_out,leaky_relu,benchmark/test_leaky_relu.py,,yes,leaky_relu_out
+main,lerp_tensor,lerp,benchmark/test_lerp.py,,yes,lerp_tensor
+main,lerp_tensor_inplace,lerp,benchmark/test_lerp.py,,yes,lerp_tensor_
+main,lerp_scalar,lerp,benchmark/test_lerp.py,,yes,lerp_scalar
+main,lerp_scalar_inplace,lerp,benchmark/test_lerp.py,,yes,lerp_scalar_
+main,less,less,benchmark/test_less.py,,yes,less
+main,less_scalar,less,benchmark/test_less.py,,yes,less_scalar
+main,less_,less_,benchmark/test_less_.py,,yes,less_
+main,less_scalar_,less_,benchmark/test_less_.py,,yes,less_scalar_
+main,less_equal,less_equal,benchmark/test_less_equal.py,,yes,less_equal
+main,less_equal_scalar,less_equal,benchmark/test_less_equal.py,,yes,less_equal_scalar
+main,less_equal_,less_equal_,benchmark/test_less_equal_.py,,yes,less_equal_
+main,less_equal_scalar_,less_equal_,benchmark/test_less_equal_.py,,yes,less_equal_scalar_
+main,lgamma,lgamma_,benchmark/test_lgamma_.py,,yes,lgamma
+main,lgamma_,lgamma_,benchmark/test_lgamma_.py,,yes,lgamma_
+main,lift,lift,benchmark/test_lift.py,,yes,lift
+main,lift_out,lift,benchmark/test_lift.py,,yes,lift_out
+main,lift_fresh,lift_fresh,benchmark/test_lift_fresh.py,,yes,lift_fresh
+main,lift_fresh_copy,lift_fresh_copy,benchmark/test_lift_fresh_copy.py,,yes,lift_fresh_copy
+main,linalg_cholesky,linalg_cholesky,benchmark/test_linalg_cholesky.py,,yes,linalg_cholesky
+main,linalg_eigvals,linalg_eigvals,benchmark/test_linalg_eigvals.py,,yes,linalg_eigvals
+main,linalg_ldl_factor,linalg_ldl_factor,benchmark/test_linalg_ldl_factor.py,,yes,linalg_ldl_factor
+main,linalg_ldl_factor_ex,linalg_ldl_factor_ex,benchmark/test_linalg_ldl_factor_ex.py,,yes,linalg_ldl_factor_ex
+main,linalg_ldl_solve,linalg_ldl_solve,benchmark/test_linalg_ldl_solve.py,,yes,linalg_ldl_solve
+main,linalg_lstsq,linalg_lstsq,benchmark/test_linalg_lstsq.py,,yes,linalg_lstsq
+main,linalg_lu_factor,linalg_lu_factor,benchmark/test_linalg_lu_factor.py,,yes,linalg_lu_factor
+main,linalg_lu_factor_out,linalg_lu_factor,benchmark/test_linalg_lu_factor.py,,yes,linalg_lu_factor_out
+main,linalg_lu_factor_ex,linalg_lu_factor_ex,benchmark/test_linalg_lu_factor_ex.py,,yes,linalg_lu_factor_ex
+main,linalg_lu_factor_ex_out,linalg_lu_factor_ex,benchmark/test_linalg_lu_factor_ex.py,,yes,linalg_lu_factor_ex_out
+main,linalg_slogdet,linalg_slogdet,benchmark/test_linalg_slogdet.py,,yes,linalg_slogdet
+main,linalg_svdvals,linalg_svdvals,benchmark/test_linalg_svdvals.py,,yes,linalg_svdvals
+main,linear,linear,benchmark/test_linear.py,,yes,linear
+main,linear_backward,linear_backward,benchmark/test_linear_backward.py,,yes,linear_backward
+main,linspace,linspace,benchmark/test_linspace.py,,yes,linspace
+main,log,log,benchmark/test_log.py,,yes,log
+main,log10,log10,benchmark/test_log10.py,,yes,log10
+main,log10_inplace,log10,benchmark/test_log10.py,,yes,log10_
+main,log10_out,log10,benchmark/test_log10.py,,yes,log10_out
+main,log1p,log1p,benchmark/test_log1p.py,,yes,log1p
+main,log1p_inplace,log1p,benchmark/test_log1p.py,,yes,log1p_
+main,log2,log2,benchmark/test_log2.py,,yes,log2
+main,log2_,log2,benchmark/test_log2.py,,yes,log2_
+main,log_inplace,log_,benchmark/test_log_.py,,yes,log_
+main,log_normal_,log_normal_,benchmark/test_log_normal_.py,,yes,log_normal_
+main,log_sigmoid,log_sigmoid,benchmark/test_log_sigmoid.py,,yes,log_sigmoid
+main,log_sigmoid_forward,log_sigmoid_forward,benchmark/test_log_sigmoid_forward.py,,yes,log_sigmoid_forward
+main,log_softmax,log_softmax,benchmark/test_log_softmax.py,,yes,log_softmax
+main,log_softmax_out,log_softmax,benchmark/test_log_softmax.py,,yes,log_softmax_out
+main,log_softmax_backward_data,log_softmax,benchmark/test_log_softmax.py,,yes,log_softmax_backward_data
+main,log_softmax_backward_data_out,log_softmax,benchmark/test_log_softmax.py,,yes,log_softmax_backward_data_out
+main,logaddexp,logaddexp,benchmark/test_logaddexp.py,,yes,logaddexp
+main,logaddexp_out,logaddexp,benchmark/test_logaddexp.py,,yes,logaddexp_out
+main,logaddexp2,logaddexp2,benchmark/test_logaddexp2.py,,yes,logaddexp2
+main,logaddexp2_out,logaddexp2,benchmark/test_logaddexp2.py,,yes,logaddexp2_out
+main,logcumsumexp,logcumsumexp,benchmark/test_logcumsumexp.py,,yes,logcumsumexp
+main,logcumsumexp_out,logcumsumexp,benchmark/test_logcumsumexp.py,,yes,logcumsumexp_out
+main,logical_and,logical_and,benchmark/test_logical_and.py,,yes,logical_and
+main,logical_and_inplace,logical_and,benchmark/test_logical_and.py,,yes,logical_and_
+main,logical_not,logical_not,benchmark/test_logical_not.py,,yes,logical_not
+main,logical_not_,logical_not,benchmark/test_logical_not.py,,yes,logical_not_
+main,logical_or,logical_or,benchmark/test_logical_or.py,,yes,logical_or
+main,logical_or_inplace,logical_or,benchmark/test_logical_or.py,,yes,logical_or_
+main,logical_xor,logical_xor,benchmark/test_logical_xor.py,,yes,logical_xor
+main,logical_xor_,logical_xor_,benchmark/test_logical_xor_.py,,yes,logical_xor_
+main,logit,logit,benchmark/test_logit.py,,yes,logit
+main,logit_inplace,logit,benchmark/test_logit.py,,yes,logit_
+main,logit_out,logit,benchmark/test_logit.py,,yes,logit_out
+main,logspace,logspace,benchmark/test_logspace.py,,yes,logspace
+main,logsumexp,logsumexp,benchmark/test_logsumexp.py,,yes,logsumexp
+main,lshift,lshift,benchmark/test_lshift.py,,yes,lshift
+main,lstm,lstm,benchmark/test_lstm.py,,yes,lstm
+main,lt,lt,benchmark/test_lt.py,,yes,lt
+main,lt_scalar,lt,benchmark/test_lt.py,,yes,lt_scalar
+main,lt_,lt_,benchmark/test_lt_.py,,yes,lt_
+main,lt_scalar_,lt_,benchmark/test_lt_.py,,yes,lt_scalar_
+main,lu_unpack,lu_unpack,benchmark/test_lu_unpack.py,,yes,lu_unpack
+main,make_dep_token,make_dep_token,benchmark/test_make_dep_token.py,,yes,make_dep_token
+main,margin_ranking_loss,margin_ranking_loss,benchmark/test_margin_ranking_loss.py,,yes,margin_ranking_loss
+main,margin_ranking_loss_backward,margin_ranking_loss,benchmark/test_margin_ranking_loss.py,,yes,margin_ranking_loss
+main,masked_fill,masked_fill,benchmark/test_masked_fill.py,,yes,masked_fill
+main,masked_fill_inplace,masked_fill,benchmark/test_masked_fill.py,,yes,masked_fill_
+main,masked_fill_scalar,masked_fill,benchmark/test_masked_fill.py,,yes,masked_fill_scalar
+main,masked_fill_scalar_inplace,masked_fill,benchmark/test_masked_fill.py,,yes,masked_fill_scalar_
+main,masked_scale,masked_scale,benchmark/test_masked_scale.py,,yes,masked_scale
+main,masked_scatter,masked_scatter,benchmark/test_masked_scatter.py,,yes,masked_scatter
+main,masked_scatter_inplace,masked_scatter,benchmark/test_masked_scatter.py,,yes,masked_scatter_
+main,masked_select,masked_select,benchmark/test_masked_select.py,,yes,masked_select
+main,matmul_bias_activation,matmul_bias_activation,benchmark/test_matmul_bias_activation.py,,yes,matmul_bias_activation
+main,matmuladd,matmuladd,benchmark/test_matmuladd.py,,yes,matmuladd
+main,max,max,benchmark/test_max.py,,yes,max
+main,max_dim,max,benchmark/test_max.py,,yes,max_dim
+main,max_pool2d_with_indices,max_pool2d,benchmark/test_max_pool2d.py,,yes,max_pool2d_with_indices
+main,max_pool2d_backward,max_pool2d,benchmark/test_max_pool2d.py,,yes,max_pool2d_backward
+main,max_pool2d_with_indices_backward,max_pool2d,benchmark/test_max_pool2d.py,,yes,max_pool2d_with_indices_backward
+main,perf_max_pool3d,max_pool3d,benchmark/test_max_pool3d.py,,yes,max_pool3d
+main,perf_max_pool3d_backward,max_pool3d,benchmark/test_max_pool3d.py,,yes,max_pool3d_backward
+main,max_unpool2d,max_unpool2d,benchmark/test_max_unpool2d.py,,yes,max_unpool2d
+main,max_unpool3d,max_unpool3d,benchmark/test_max_unpool3d.py,,yes,max_unpool3d
+main,maximum,maximum,benchmark/test_maximum.py,,yes,maximum
+main,mean,mean,benchmark/test_mean.py,,yes,mean
+main,mean_dim,mean,benchmark/test_mean.py,,yes,mean_dim
+main,median,median,benchmark/test_median.py,,yes,median
+main,median_dim,median,benchmark/test_median.py,,yes,median_dim
+main,mhc_post,mhc,benchmark/test_mhc.py,,yes,mhc_post
+main,mhc_pre,mhc,benchmark/test_mhc.py,,yes,mhc_pre
+main,hc_split_sinkhorn_forward,mhc,benchmark/test_mhc.py,,yes,hc_split_sinkhorn_forward
+main,mhc_bwd,mhc,benchmark/test_mhc.py,,yes,mhc_bwd
+main,hc_head_fused_kernel,mhc,benchmark/test_mhc.py,,yes,hc_head_fused_kernel
+main,min,min,benchmark/test_min.py,,yes,min
+main,min_dim,min,benchmark/test_min.py,,yes,min_dim
+main,minimum,minimum,benchmark/test_minimum.py,,yes,minimum
+main,miopen_batch_norm_backward,miopen_batch_norm_backward,benchmark/test_miopen_batch_norm_backward.py,,yes,miopen_batch_norm_backward
+main,mish,mish,benchmark/test_mish.py,,yes,mish
+main,mish_inplace,mish,benchmark/test_mish.py,,yes,mish_
+main,mish_backward,mish,benchmark/test_mish.py,,yes,mish_backward
+main,mm,mm,benchmark/test_mm.py,,yes,mm
+main,mm_self_transpose_benchmark,mm,benchmark/test_mm.py,,yes,mm_self_transpose
+main,mm_out,mm,benchmark/test_mm.py,,yes,mm_out
+main,perf_mode,mode,benchmark/test_mode.py,,yes,mode
+main,moe_align_block_size_triton,moe_align_block_size_triton,benchmark/test_moe_align_block_size_triton.py,,yes,moe_align_block_size_triton
+main,moe_align_block_size_fast_paths,moe_align_block_size_triton,benchmark/test_moe_align_block_size_triton.py,,yes,moe_align_block_size_fast_paths
+main,moe_sum,moe_sum,benchmark/test_moe_sum.py,,yes,moe_sum
+main,mrope_gems_vs_torch,mrope,benchmark/test_mrope.py,,yes,mrope
+main,mse_loss,mse_loss,benchmark/test_mse_loss.py,,yes,mse_loss
+main,mse_loss_backward,mse_loss_backward,benchmark/test_mse_loss_backward.py,,yes,mse_loss_backward
+main,mul,mul,benchmark/test_mul.py,,yes,mul
+main,mul_inplace,mul,benchmark/test_mul.py,,yes,mul_
+main,multinomial_with_replacement,multinomial,benchmark/test_multinomial.py,,yes,multinomial
+main,multiply,multiply,benchmark/test_multiply.py,,yes,multiply
+main,multiply_,multiply_,benchmark/test_multiply_.py,,yes,multiply_
+main,mv,mv,benchmark/test_mv.py,,yes,mv
+main,mvlgamma,mvlgamma,benchmark/test_mvlgamma.py,,yes,mvlgamma
+main,mvlgamma_,mvlgamma_,benchmark/test_mvlgamma_.py,,yes,mvlgamma_
+main,nan_to_num,nan_to_num,benchmark/test_nan_to_num.py,,yes,nan_to_num
+main,nan_to_num_,nan_to_num_,benchmark/test_nan_to_num_.py,,yes,nan_to_num_
+main,nanmedian,nanmedian,benchmark/test_nanmedian.py,,yes,nanmedian
+main,nanmedian_out,nanmedian,benchmark/test_nanmedian.py,,yes,nanmedian_out
+main,nanmedian_dim,nanmedian,benchmark/test_nanmedian.py,,yes,nanmedian_dim
+main,nanmedian_dim_values,nanmedian,benchmark/test_nanmedian.py,,yes,nanmedian_dim_values
+main,benchmark_nansum,nansum,benchmark/test_nansum.py,,yes,nansum
+main,narrow,narrow,benchmark/test_narrow.py,,yes,narrow
+main,narrow_copy_perf,narrow_copy,benchmark/test_narrow_copy.py,,yes,narrow_copy
+main,native_batch_norm_legit_functional,native_batch_norm_legit_functional,benchmark/test_native_batch_norm_legit_functional.py,,yes,native_batch_norm_legit_functional
+main,ne,ne,benchmark/test_ne.py,,yes,ne
+main,ne_scalar,ne,benchmark/test_ne.py,,yes,ne_scalar
+main,ne_,ne_,benchmark/test_ne_.py,,yes,ne_
+main,ne_scalar_,ne_,benchmark/test_ne_.py,,yes,ne_scalar_
+main,neg,neg,benchmark/test_neg.py,,yes,neg
+main,neg_inplace,neg,benchmark/test_neg.py,,yes,neg_
+main,negative,negative,benchmark/test_negative.py,,yes,negative
+main,negative_,negative_,benchmark/test_negative_.py,,yes,negative_
+main,nested_view_from_buffer_copy,nested_view_from_buffer_copy,benchmark/test_nested_view_from_buffer_copy.py,,yes,nested_view_from_buffer_copy
+main,new_full,new_full,benchmark/test_new_full.py,,yes,new_full
+main,new_ones,new_ones,benchmark/test_new_ones.py,,yes,new_ones
+main,nextafter,nextafter,benchmark/test_nextafter.py,,yes,nextafter
+main,nextafter_,nextafter,benchmark/test_nextafter.py,,yes,nextafter_
+main,nextafter_,nextafter_,benchmark/test_nextafter_.py,,yes,nextafter_
+main,nll_loss_forward,nll_loss,benchmark/test_nll_loss.py,,yes,nll_loss_forward
+main,nll_loss_backward,nll_loss,benchmark/test_nll_loss.py,,yes,nll_loss_backward
+main,nll_loss2d_forward,nll_loss2d,benchmark/test_nll_loss2d.py,,yes,nll_loss2d_forward
+main,nll_loss2d_backward,nll_loss2d,benchmark/test_nll_loss2d.py,,yes,nll_loss2d_backward
+main,nll_loss_nd_forward,nll_loss_nd,benchmark/test_nll_loss_nd.py,,yes,nll_loss_nd_forward
+main,nll_loss_nd_backward,nll_loss_nd,benchmark/test_nll_loss_nd.py,,yes,nll_loss_nd_backward
+main,nonzero,nonzero,benchmark/test_nonzero.py,,yes,nonzero
+main,nonzero_numpy,nonzero_numpy,benchmark/test_nonzero_numpy.py,,yes,nonzero_numpy
+main,perf_nonzero_static,nonzero_static,benchmark/test_nonzero_static.py,,yes,nonzero_static
+main,norm,norm,benchmark/test_norm.py,,yes,norm
+main,norm_scalar,norm,benchmark/test_norm.py,,yes,norm_scalar
+main,norm_scalaropt_dim,norm,benchmark/test_norm.py,,yes,norm_scalaropt_dim
+main,normal_tensor_tensor,normal,benchmark/test_normal.py,,yes,normal_tensor_tensor
+main,normal_tensor_float,normal,benchmark/test_normal.py,,yes,normal_tensor_float
+main,normal_inplace,normal,benchmark/test_normal.py,,yes,normal_float_float_
+main,normal_float_tensor,normal,benchmark/test_normal.py,,yes,normal_float_tensor
+main,normed_cumsum,normed_cumsum,benchmark/test_normed_cumsum.py,,yes,normed_cumsum
+main,not_equal,not_equal,benchmark/test_not_equal.py,,yes,not_equal
+main,not_equal_scalar,not_equal,benchmark/test_not_equal.py,,yes,not_equal_scalar
+main,not_equal_,not_equal_,benchmark/test_not_equal_.py,,yes,not_equal_
+main,not_equal_scalar_,not_equal_,benchmark/test_not_equal_.py,,yes,not_equal_scalar_
+main,one_hot,one_hot,benchmark/test_one_hot.py,,yes,one_hot
+main,ones,ones,benchmark/test_ones.py,,yes,ones
+main,ones_like,ones_like,benchmark/test_ones_like.py,,yes,ones_like
+main,outer,outer,benchmark/test_outer.py,,yes,outer
+main,outplace_fused_experts_gems_vs_vllm,outplace_fused_experts,benchmark/test_outplace_fused_experts.py,,yes,outplace_fused_experts
+main,pack_seq,pack_seq,benchmark/test_pack_seq.py,,yes,pack_seq_triton
+main,pack_seq_fp8,pack_seq,benchmark/test_pack_seq.py,,yes,pack_seq_triton
+main,pad,pad,benchmark/test_pad.py,,yes,pad
+main,pairwise_distance,pairwise_distance,benchmark/test_pairwise_distance.py,,yes,pairwise_distance
+main,pdist,pdist,benchmark/test_pdist.py,,yes,pdist
+main,pdist_backward,pdist_backward,benchmark/test_pdist_backward.py,,yes,pdist_backward
+main,pdist_forward,pdist_forward_perf,benchmark/test_pdist_forward_perf.py,,yes,pdist_forward
+main,per_token_group_quant_fp8,per_token_group_quant_fp8,benchmark/test_per_token_group_quant_fp8.py,,yes,per_token_group_quant_fp8
+main,permute_copy,permute_copy,benchmark/test_permute_copy.py,,yes,permute_copy
+main,pixel_shuffle,pixel_shuffle,benchmark/test_pixel_shuffle.py,,yes,pixel_shuffle
+main,pixel_shuffle_out,pixel_shuffle,benchmark/test_pixel_shuffle.py,,yes,pixel_shuffle_out
+main,pixel_unshuffle,pixel_unshuffle,benchmark/test_pixel_unshuffle.py,,yes,pixel_unshuffle
+main,pixel_unshuffle_out,pixel_unshuffle,benchmark/test_pixel_unshuffle.py,,yes,pixel_unshuffle_out
+main,poisson,poisson,benchmark/test_poisson.py,,yes,poisson
+main,polar,polar,benchmark/test_polar.py,,yes,polar
+main,polygamma,polygamma,benchmark/test_polygamma.py,,yes,polygamma
+main,polygamma_inplace,polygamma,benchmark/test_polygamma.py,,yes,polygamma_
+main,polygamma_out,polygamma,benchmark/test_polygamma.py,,yes,polygamma_out
+main,pow_tensor_tensor,pow,benchmark/test_pow.py,,yes,pow_tensor_tensor
+main,pow_inplace,pow,benchmark/test_pow.py,,yes,pow_tensor_tensor_
+main,pow_scalar,pow,benchmark/test_pow.py,,yes,pow_scalar
+main,pow_tensor_scalar,pow,benchmark/test_pow.py,,yes,pow_tensor_scalar
+main,pow_tensor_scalar_inplace,pow,benchmark/test_pow.py,,yes,pow_tensor_scalar_
+main,prelu,prelu,benchmark/test_prelu.py,,yes,prelu
+main,prelu_kernel,prelu_kernel,benchmark/test_prelu_kernel.py,,yes,prelu_kernel
+main,prelu_kernel_backward,prelu_kernel_backward,benchmark/test_prelu_kernel_backward.py,,yes,prelu_kernel_backward
+main,prod,prod,benchmark/test_prod.py,,yes,prod
+main,quantile,quantile,benchmark/test_quantile.py,,yes,quantile
+main,rad2deg,rad2deg,benchmark/test_rad2deg.py,,yes,rad2deg
+main,rad2deg_,rad2deg,benchmark/test_rad2deg.py,,yes,rad2deg_
+main,rand,rand,benchmark/test_rand.py,,yes,rand
+main,rand_like,rand_like,benchmark/test_rand_like.py,,yes,rand_like
+main,randint,randint,benchmark/test_randint.py,,yes,randint
+main,randint_like,randint_like,benchmark/test_randint_like.py,,yes,randint_like
+main,randn,randn,benchmark/test_randn.py,,yes,randn
+main,randn_like,randn_like,benchmark/test_randn_like.py,,yes,randn_like
+main,randperm,randperm,benchmark/test_randperm.py,,yes,randperm
+main,range,range,benchmark/test_range.py,,yes,range
+main,reciprocal,reciprocal,benchmark/test_reciprocal.py,,yes,reciprocal
+main,reciprocal_inplace,reciprocal,benchmark/test_reciprocal.py,,yes,reciprocal_
+main,reflection_pad1d,reflection_pad1d,benchmark/test_reflection_pad1d.py,,yes,reflection_pad1d
+main,reflection_pad1d_out,reflection_pad1d,benchmark/test_reflection_pad1d.py,,yes,reflection_pad1d_out
+main,reflection_pad1d_backward,reflection_pad1d_backward,benchmark/test_reflection_pad1d_backward.py,,yes,reflection_pad1d_backward
+main,reflection_pad2d,reflection_pad2d,benchmark/test_reflection_pad2d.py,,yes,reflection_pad2d
+main,reflection_pad2d_out,reflection_pad2d,benchmark/test_reflection_pad2d.py,,yes,reflection_pad2d_out
+main,reflection_pad2d_backward,reflection_pad2d_backward,benchmark/test_reflection_pad2d_backward.py,,yes,reflection_pad2d_backward
+main,reflection_pad3d,reflection_pad3d,benchmark/test_reflection_pad3d.py,,yes,reflection_pad3d
+main,reflection_pad3d_out,reflection_pad3d,benchmark/test_reflection_pad3d.py,,yes,reflection_pad3d_out
+main,reflection_pad3d_backward,reflection_pad3d_backward,benchmark/test_reflection_pad3d_backward.py,,yes,reflection_pad3d_backward
+main,reglu,reglu,benchmark/test_reglu.py,,yes,reglu
+main,relu,relu,benchmark/test_relu.py,,yes,relu
+main,relu_inplace,relu,benchmark/test_relu.py,,yes,relu_
+main,relu6,relu6,benchmark/test_relu6.py,,yes,relu6
+main,remainder,remainder,benchmark/test_remainder.py,,yes,remainder
+main,remainder_tensor,remainder,benchmark/test_remainder.py,,yes,remainder_tensor
+main,remainder_tensor_inplace,remainder,benchmark/test_remainder.py,,yes,remainder_tensor_
+main,remainder_scalar,remainder,benchmark/test_remainder.py,,yes,remainder_scalar
+main,remainder_scalar_inplace,remainder,benchmark/test_remainder.py,,yes,remainder_scalar_
+main,remainder_scalar_tensor,remainder,benchmark/test_remainder.py,,yes,remainder_scalar_tensor
+main,renorm,renorm,benchmark/test_renorm.py,,yes,renorm
+main,renorm_,renorm_,benchmark/test_renorm_.py,,yes,renorm_
+main,repeat,repeat,benchmark/test_repeat.py,,yes,repeat
+main,repeat_interleave_self_int,repeat_interleave,benchmark/test_repeat_interleave.py,,yes,repeat_interleave_self_int
+main,repeat_interleave_self_tensor,repeat_interleave,benchmark/test_repeat_interleave.py,,yes,repeat_interleave_self_tensor
+main,repeat_interleave_tensor,repeat_interleave,benchmark/test_repeat_interleave.py,,yes,repeat_interleave_tensor
+main,replication_pad1d,replication_pad1d,benchmark/test_replication_pad1d.py,,yes,replication_pad1d
+main,replication_pad2d,replication_pad2d,benchmark/test_replication_pad2d.py,,yes,replication_pad2d
+main,replication_pad2d_out,replication_pad2d,benchmark/test_replication_pad2d.py,,yes,replication_pad2d_out
+main,replication_pad2d_backward,replication_pad2d_backward,benchmark/test_replication_pad2d_backward.py,,yes,replication_pad2d_backward
+main,replication_pad3d,replication_pad3d,benchmark/test_replication_pad3d.py,,yes,replication_pad3d
+main,replication_pad3d_backward,replication_pad3d_backward,benchmark/test_replication_pad3d_backward.py,,yes,replication_pad3d_backward
+main,reshape_and_cache,reshape_and_cache,benchmark/test_reshape_and_cache.py,,yes,reshape_and_cache
+main,reshape_and_cache_flash,reshape_and_cache_flash,benchmark/test_reshape_and_cache_flash.py,,yes,reshape_and_cache_flash
+main,resize,resize,benchmark/test_resize.py,,yes,resize
+main,resize_,resize,benchmark/test_resize.py,,yes,resize_
+main,resize_as,resize_as,benchmark/test_resize_as.py,,yes,resize_as
+main,resize_as_,resize_as,benchmark/test_resize_as.py,,yes,resize_as_
+main,resize_output,resize_output,benchmark/test_resize_output.py,,yes,resize_output
+main,resolve_conj,resolve_conj,benchmark/test_resolve_conj.py,,yes,resolve_conj
+main,resolve_neg,resolve_neg,benchmark/test_resolve_neg.py,,yes,resolve_neg
+main,rms_norm,rms_norm,benchmark/test_rms_norm.py,,yes,rms_norm
+main,rnn_relu,rnn_relu,benchmark/test_rnn_relu.py,,yes,rnn_relu
+main,roll,roll,benchmark/test_roll.py,,yes,roll
+main,rot90,rot90,benchmark/test_rot90.py,,yes,rot90
+main,round,round,benchmark/test_round.py,,yes,round
+main,round_inplace,round,benchmark/test_round.py,,yes,round_
+main,round_out,round,benchmark/test_round.py,,yes,round_out
+main,perf_router_gemm,router_gemm,benchmark/test_router_gemm.py,,yes,router_gemm
+main,rrelu_with_noise_backward,rrelu_with_noise_backward,benchmark/test_rrelu_with_noise_backward.py,,yes,rrelu_with_noise_backward
+main,rrelu_with_noise_functional,rrelu_with_noise_functional,benchmark/test_rrelu_with_noise_functional.py,,yes,rrelu_with_noise_functional
+main,rsqrt,rsqrt,benchmark/test_rsqrt.py,,yes,rsqrt
+main,rsqrt_inplace,rsqrt,benchmark/test_rsqrt.py,,yes,rsqrt_
+main,rsub_tensor,rsub,benchmark/test_rsub.py,,yes,rsub_tensor
+main,rsub_scalar,rsub,benchmark/test_rsub.py,,yes,rsub_scalar
+main,rwkv_ka_fusion,rwkv_ka_fusion,benchmark/test_rwkv_ka_fusion.py,,yes,rwkv_ka_fusion
+main,rwkv_mm_sparsity,rwkv_mm_sparsity,benchmark/test_rwkv_mm_sparsity.py,,yes,rwkv_mm_sparsity
+main,safe_softmax,safe_softmax,benchmark/test_safe_softmax.py,,yes,safe_softmax
+main,scalar_tensor,scalar_tensor,benchmark/test_scalar_tensor.py,,yes,scalar_tensor
+main,scaled_dot_product_flash_attention,scaled_dot_product_attention,benchmark/test_scaled_dot_product_attention.py,,yes,scaled_dot_product_flash_attention
+main,scaled_dot_product_attention,scaled_dot_product_attention,benchmark/test_scaled_dot_product_attention.py,,yes,scaled_dot_product_attention
+main,scaled_dot_product_fused_attention_overrideable,scaled_dot_product_fused_attention_overrideable,benchmark/test_scaled_dot_product_fused_attention_overrideable.py,,yes,scaled_dot_product_fused_attention_overrideable
+main,scaled_grouped_mm_benchmark,scaled_grouped_mm,benchmark/test_scaled_grouped_mm.py,,yes,
+main,scaled_mm_benchmark,scaled_mm,benchmark/test_scaled_mm.py,,yes,scaled_mm
+main,scaled_mm_out_benchmark,scaled_mm,benchmark/test_scaled_mm.py,,yes,scaled_mm_out
+main,scaled_softmax_forward,scaled_softmax,benchmark/test_scaled_softmax.py,,yes,scaled_softmax_forward
+main,perf_scaled_softmax_backward,scaled_softmax,benchmark/test_scaled_softmax.py,,yes,scaled_softmax_backward
+main,scatter_add_inplace,scatter_add,benchmark/test_scatter_add.py,,yes,scatter_add_
+main,scatter_add,scatter_add,benchmark/test_scatter_add.py,,yes,scatter_add
+main,scatter_reduce_add,scatter_reduce,benchmark/test_scatter_reduce.py,,yes,scatter_reduce
+main,scatter_reduce_multiply,scatter_reduce,benchmark/test_scatter_reduce.py,,yes,scatter_reduce
+main,scatter_reduce_add_inplace,scatter_reduce,benchmark/test_scatter_reduce.py,,yes,scatter_reduce_
+main,scatter_reduce_multiply_inplace,scatter_reduce,benchmark/test_scatter_reduce.py,,yes,scatter_reduce_
+main,scatter_reduce_two_sum,scatter_reduce,benchmark/test_scatter_reduce.py,,yes,scatter_reduce.two
+main,scatter_reduce_two_amax,scatter_reduce,benchmark/test_scatter_reduce.py,,yes,scatter_reduce.two
+main,scatter_reduce_two_inplace_sum,scatter_reduce_two,benchmark/test_scatter_reduce_two.py,,yes,scatter_reduce_
+main,scatter_reduce_two_inplace_amax,scatter_reduce_two,benchmark/test_scatter_reduce_two.py,,yes,scatter_reduce_
+main,scatter_reduce_two_inplace_amin,scatter_reduce_two,benchmark/test_scatter_reduce_two.py,,yes,scatter_reduce_
+main,scatter_reduce_two_inplace_mean,scatter_reduce_two,benchmark/test_scatter_reduce_two.py,,yes,scatter_reduce_
+main,scatter_src,scatter_src,benchmark/test_scatter_src.py,,yes,scatter_src
+main,scatter_src_inplace,scatter_src,benchmark/test_scatter_src.py,,yes,scatter_src_
+main,searchsorted,searchsorted,benchmark/test_searchsorted.py,,yes,searchsorted
+main,searchsorted_scalar,searchsorted,benchmark/test_searchsorted.py,,yes,searchsorted_scalar
+main,searchsorted_out,searchsorted,benchmark/test_searchsorted.py,,yes,searchsorted_out
+main,searchsorted_scalar_out,searchsorted,benchmark/test_searchsorted.py,,yes,searchsorted_scalar_out
+main,segment_reduce,segment_reduce,benchmark/test_segment_reduce.py,,yes,segment_reduce
+main,segment_reduce_out,segment_reduce,benchmark/test_segment_reduce.py,,yes,segment_reduce_out
+main,segment_reduce_backward,segment_reduce,benchmark/test_segment_reduce.py,,yes,segment_reduce_backward
+main,segment_reduce_backward_out,segment_reduce,benchmark/test_segment_reduce.py,,yes,segment_reduce_backward_out
+main,select_backward,select_backward,benchmark/test_select_backward.py,,yes,select_backward
+main,select_scatter,select_scatter,benchmark/test_select_scatter.py,,yes,select_scatter
+main,selu,selu,benchmark/test_selu.py,,yes,selu
+main,selu_inplace,selu,benchmark/test_selu.py,,yes,selu_
+main,sgn,sgn,benchmark/test_sgn.py,,yes,sgn
+main,sgn_inplace,sgn,benchmark/test_sgn.py,,yes,sgn_
+main,sgn_out,sgn,benchmark/test_sgn.py,,yes,sgn_out
+main,sigmoid,sigmoid,benchmark/test_sigmoid.py,,yes,sigmoid
+main,sigmoid_inplace,sigmoid,benchmark/test_sigmoid.py,,yes,sigmoid_
+main,sigmoid_backward,sigmoid,benchmark/test_sigmoid.py,,yes,sigmoid_backward
+main,sign,sign,benchmark/test_sign.py,,yes,sign
+main,sign_out,sign,benchmark/test_sign.py,,yes,sign_out
+main,signbit,signbit,benchmark/test_signbit.py,,yes,signbit
+main,signbit_out,signbit,benchmark/test_signbit.py,yes,yes,signbit_out
+main,silu,silu,benchmark/test_silu.py,,yes,silu
+main,silu_inplace,silu,benchmark/test_silu.py,,yes,silu_
+main,silu_backward,silu,benchmark/test_silu.py,,yes,silu_backward
+main,silu_and_mul,silu_and_mul,benchmark/test_silu_and_mul.py,,yes,silu_and_mul
+main,silu_and_mul_out,silu_and_mul,benchmark/test_silu_and_mul.py,,yes,silu_and_mul_out
+main,silu_and_mul_with_clamp,silu_and_mul_with_clamp,benchmark/test_silu_and_mul_with_clamp.py,,yes,silu_and_mul_with_clamp
+main,silu_and_mul_with_clamp_out,silu_and_mul_with_clamp,benchmark/test_silu_and_mul_with_clamp.py,,yes,silu_and_mul_with_clamp_out
+main,sin,sin,benchmark/test_sin.py,,yes,sin
+main,sin_inplace,sin,benchmark/test_sin.py,,yes,sin_
+main,sinc,sinc,benchmark/test_sinc.py,,yes,sinc
+main,sinc_,sinc,benchmark/test_sinc.py,,yes,sinc_
+main,sinh,sinh,benchmark/test_sinh.py,,yes,sinh
+main,sinh_inplace,sinh,benchmark/test_sinh.py,,yes,sinh_
+main,skip_layernorm,skip_layer_norm,benchmark/test_skip_layer_norm.py,,yes,skip_layer_norm
+main,slice_backward,slice_backward,benchmark/test_slice_backward.py,,yes,slice_backward
+main,slice_scatter,slice_scatter,benchmark/test_slice_scatter.py,,yes,slice_scatter
+main,smooth_l1_loss,smooth_l1_loss,benchmark/test_smooth_l1_loss.py,,yes,smooth_l1_loss
+main,smooth_l1_loss_backward,smooth_l1_loss,benchmark/test_smooth_l1_loss.py,,yes,smooth_l1_loss_backward
+main,soft_margin_loss,soft_margin_loss,benchmark/test_soft_margin_loss.py,,yes,soft_margin_loss
+main,soft_margin_loss_backward,soft_margin_loss_backward,benchmark/test_soft_margin_loss_backward.py,,yes,soft_margin_loss_backward
+main,softmax,softmax,benchmark/test_softmax.py,,yes,softmax
+main,softmax_backward,softmax,benchmark/test_softmax.py,,yes,softmax
+main,softmax_backward_out,softmax,benchmark/test_softmax.py,,yes,softmax_backward_out
+main,softmax_out,softmax,benchmark/test_softmax.py,,yes,softmax_out
+main,softplus,softplus,benchmark/test_softplus.py,,yes,softplus
+main,softplus_backward,softplus,benchmark/test_softplus.py,,yes,softplus_backward
+main,softshrink,softshrink,benchmark/test_softshrink.py,,yes,softshrink
+main,softshrink_out,softshrink,benchmark/test_softshrink.py,,yes,softshrink_out
+main,perf_sort,sort,benchmark/test_sort.py,,yes,sort
+main,perf_sort_stable,sort,benchmark/test_sort.py,,yes,sort_stable
+main,sparse_attn_triton,sparse_attention,benchmark/test_sparse_attention.py,,yes,sparse_attention
+main,sparse_mla_fwd_interface,sparse_mla_fwd_interface,benchmark/test_sparse_mla_fwd_interface.py,,yes,sparse_mla_fwd_interface
+main,sparse_semi_structured_mm,sparse_semi_structured_mm,benchmark/test_sparse_semi_structured_mm.py,,yes,sparse_semi_structured_mm
+main,special_airy_ai,special_airy_ai,benchmark/test_special_airy_ai.py,,yes,special_airy_ai
+main,special_airy_ai_out,special_airy_ai,benchmark/test_special_airy_ai.py,,yes,special_airy_ai_out
+main,special_bessel_j0,special_bessel_j0,benchmark/test_special_bessel_j0.py,,yes,special_bessel_j0
+main,special_bessel_j1,special_bessel_j1,benchmark/test_special_bessel_j1.py,,yes,special_bessel_j1
+main,special_chebyshev_polynomial_u,special_chebyshev_polynomial_u,benchmark/test_special_chebyshev_polynomial_u.py,,yes,special_chebyshev_polynomial_u
+main,special_chebyshev_polynomial_v,special_chebyshev_polynomial_v,benchmark/test_special_chebyshev_polynomial_v.py,,yes,special_chebyshev_polynomial_v
+main,special_chebyshev_polynomial_w,special_chebyshev_polynomial_w,benchmark/test_special_chebyshev_polynomial_w.py,,yes,special_chebyshev_polynomial_w
+main,special_chebyshev_polynomial_w_out,special_chebyshev_polynomial_w,benchmark/test_special_chebyshev_polynomial_w.py,,yes,special_chebyshev_polynomial_w_out
+main,special_digamma,special_digamma,benchmark/test_special_digamma.py,,yes,special_digamma
+main,special_erfc,special_erfc,benchmark/test_special_erfc.py,,yes,special_erfc
+main,erfc,special_erfc,benchmark/test_special_erfc.py,,yes,erfc
+main,erfc_,special_erfc,benchmark/test_special_erfc.py,,yes,erfc_
+main,special_erfcx,special_erfcx,benchmark/test_special_erfcx.py,,yes,special_erfcx
+main,special_erfinv,special_erfinv,benchmark/test_special_erfinv.py,,yes,special_erfinv
+main,special_erfinv_out,special_erfinv,benchmark/test_special_erfinv.py,,yes,special_erfinv_out
+main,special_gammainc,special_gammainc,benchmark/test_special_gammainc.py,,yes,special_gammainc
+main,special_gammaln,special_gammaln,benchmark/test_special_gammaln.py,,yes,special_gammaln
+main,special_gammaln_out,special_gammaln,benchmark/test_special_gammaln.py,,yes,special_gammaln_out
+main,special_hermite_polynomial_h,special_hermite_polynomial_h,benchmark/test_special_hermite_polynomial_h.py,,yes,special_hermite_polynomial_h
+main,special_i0e,special_i0e,benchmark/test_special_i0e.py,,yes,special_i0e
+main,special_i0e_out,special_i0e,benchmark/test_special_i0e.py,,yes,special_i0e_out
+main,special_i1,special_i1,benchmark/test_special_i1.py,,yes,special_i1
+main,special_i1_out,special_i1,benchmark/test_special_i1.py,,yes,special_i1_out
+main,special_i1e,special_i1e,benchmark/test_special_i1e.py,,yes,special_i1e
+main,special_legendre_polynomial_p,special_legendre_polynomial_p,benchmark/test_special_legendre_polynomial_p.py,,yes,special_legendre_polynomial_p
+main,special_log1p,special_log1p,benchmark/test_special_log1p.py,,yes,special_log1p
+main,special_log1p_out,special_log1p,benchmark/test_special_log1p.py,,yes,special_log1p_out
+main,special_log1p_non_tensor,special_log1p,benchmark/test_special_log1p.py,,yes,
+main,special_log_softmax,special_log_softmax,benchmark/test_special_log_softmax.py,,yes,special_log_softmax
+main,special_logsumexp,special_logsumexp,benchmark/test_special_logsumexp.py,,yes,special_logsumexp
+main,special_modified_bessel_k0,special_modified_bessel_k0,benchmark/test_special_modified_bessel_k0.py,,yes,special_modified_bessel_k0
+main,special_modified_bessel_k0_out,special_modified_bessel_k0,benchmark/test_special_modified_bessel_k0.py,,yes,special_modified_bessel_k0_out
+main,special_modified_bessel_k1,special_modified_bessel_k1,benchmark/test_special_modified_bessel_k1.py,,yes,special_modified_bessel_k1
+main,special_modified_bessel_k1_out,special_modified_bessel_k1,benchmark/test_special_modified_bessel_k1.py,,yes,special_modified_bessel_k1_out
+main,special_round,special_round,benchmark/test_special_round.py,,yes,special_round
+main,special_round_out,special_round,benchmark/test_special_round.py,,yes,special_round_out
+main,special_scaled_modified_bessel_k1,special_scaled_modified_bessel_k1,benchmark/test_special_scaled_modified_bessel_k1.py,,yes,special_scaled_modified_bessel_k1
+main,special_scaled_modified_bessel_k1_out,special_scaled_modified_bessel_k1,benchmark/test_special_scaled_modified_bessel_k1.py,,yes,special_scaled_modified_bessel_k1_out
+main,special_shifted_chebyshev_polynomial_u,special_shifted_chebyshev_polynomial_u,benchmark/test_special_shifted_chebyshev_polynomial_u.py,,yes,special_shifted_chebyshev_polynomial_u
+main,special_shifted_chebyshev_polynomial_u_,special_shifted_chebyshev_polynomial_u,benchmark/test_special_shifted_chebyshev_polynomial_u.py,,yes,special_shifted_chebyshev_polynomial_u_
+main,special_shifted_chebyshev_polynomial_v,special_shifted_chebyshev_polynomial_v,benchmark/test_special_shifted_chebyshev_polynomial_v.py,,yes,special_shifted_chebyshev_polynomial_v
+main,special_shifted_chebyshev_polynomial_w,special_shifted_chebyshev_polynomial_w,benchmark/test_special_shifted_chebyshev_polynomial_w.py,,yes,special_shifted_chebyshev_polynomial_w
+main,special_sinc,special_sinc,benchmark/test_special_sinc.py,,yes,special_sinc
+main,special_xlog1py,special_xlog1py,benchmark/test_special_xlog1py.py,,yes,special_xlog1py
+main,split_with_sizes_copy,split_with_sizes_copy,benchmark/test_split_with_sizes_copy.py,,yes,split_with_sizes_copy
+main,sqrt,sqrt,benchmark/test_sqrt.py,,yes,sqrt
+main,sqrt_inplace,sqrt,benchmark/test_sqrt.py,,yes,sqrt_
+main,square,square,benchmark/test_square.py,,yes,square
+main,square_inplace,square,benchmark/test_square.py,,yes,square_
+main,square_out,square,benchmark/test_square.py,,yes,square_out
+main,squeeze_copy,squeeze_copy,benchmark/test_squeeze_copy.py,,yes,squeeze_copy
+main,stack,stack,benchmark/test_stack.py,,yes,stack
+main,stage_deepseek_v4_mega_moe_inputs_benchmark,stage_deepseek_v4_mega_moe_inputs,benchmark/test_stage_deepseek_v4_mega_moe_inputs.py,,yes,
+main,std,std,benchmark/test_std.py,,yes,std
+main,sub,sub,benchmark/test_sub.py,,yes,sub
+main,sub_inplace,sub,benchmark/test_sub.py,,yes,sub_
+main,subtract,subtract_,benchmark/test_subtract_.py,,yes,subtract
+main,subtract_,subtract_,benchmark/test_subtract_.py,,yes,subtract_
+main,sum,sum,benchmark/test_sum.py,,yes,sum
+main,sum_dim_out,sum,benchmark/test_sum.py,,yes,sum_dim_out
+main,sum_out,sum,benchmark/test_sum.py,,yes,sum_out
+main,sum_dim,sum_dim,benchmark/test_sum_dim.py,,yes,sum_dim
+main,svd,svd,benchmark/test_svd.py,,yes,svd
+main,swiglu,swiglu,benchmark/test_swiglu.py,,yes,swiglu
+main,sym_storage_offset,sym_storage_offset,benchmark/test_sym_storage_offset.py,,yes,sym_storage_offset
+main,sym_stride,sym_stride,benchmark/test_sym_stride.py,,yes,sym_stride
+main,t_copy,t_copy,benchmark/test_t_copy.py,,yes,t_copy
+main,t_copy_out,t_copy,benchmark/test_t_copy.py,,yes,t_copy_out
+main,take,take,benchmark/test_take.py,,yes,take
+main,take_out,take,benchmark/test_take.py,,yes,take_out
+main,tan,tan,benchmark/test_tan.py,,yes,tan
+main,tan_inplace,tan,benchmark/test_tan.py,,yes,tan_
+main,tanh,tanh,benchmark/test_tanh.py,,yes,tanh
+main,tanh_inplace,tanh,benchmark/test_tanh.py,,yes,tanh_
+main,tanh_backward,tanh,benchmark/test_tanh.py,,yes,tanh_backward
+main,te_rmsnorm_fwd,te_rmsnorm,benchmark/test_te_rmsnorm.py,,yes,te_rmsnorm_fwd
+main,te_rmsnorm_bwd,te_rmsnorm,benchmark/test_te_rmsnorm.py,,yes,te_rmsnorm_bwd
+main,tensor_split,tensor_split,benchmark/test_tensor_split.py,,yes,tensor_split
+main,thnn_fused_lstm_cell,thnn_fused_lstm_cell,benchmark/test_thnn_fused_lstm_cell.py,,yes,thnn_fused_lstm_cell
+main,thnn_fused_lstm_cell_backward_impl,thnn_fused_lstm_cell_backward_impl,benchmark/test_thnn_fused_lstm_cell_backward_impl.py,,yes,thnn_fused_lstm_cell_backward_impl
+main,threshold,threshold,benchmark/test_threshold.py,,yes,threshold
+main,threshold_backward,threshold,benchmark/test_threshold.py,,yes,threshold_backward
+main,threshold_,threshold_,benchmark/test_threshold_.py,,yes,threshold_
+main,tile,tile,benchmark/test_tile.py,,yes,tile
+main,to_copy,to_copy,benchmark/test_to_copy.py,,yes,
+main,top_k_per_row_decode,top_k_per_row_decode,benchmark/test_top_k_per_row_decode.py,,yes,top_k_per_row_decode
+main,top_k_per_row_prefill,top_k_per_row_prefill,benchmark/test_top_k_per_row_prefill.py,,yes,top_k_per_row_prefill
+main,topk,topk,benchmark/test_topk.py,,yes,topk
+main,topk_softmax,topk_softmax,benchmark/test_topk_softmax.py,,yes,topk_softmax
+main,topk_softplus_sqrt,topk_softplus_sqrt,benchmark/test_topk_softplus_sqrt.py,,yes,topk_softplus_sqrt
+main,trace,trace,benchmark/test_trace.py,,yes,trace
+main,transpose,transpose,benchmark/test_transpose.py,,yes,transpose
+main,tril,tril,benchmark/test_tril.py,,yes,tril
+main,tril_extreme_diagonal,tril,benchmark/test_tril.py,,yes,tril
+main,tril_out_transposed,tril,benchmark/test_tril.py,,yes,tril_out
+main,tril_out_sliced,tril,benchmark/test_tril.py,,yes,tril_out
+main,tril_inplace,tril,benchmark/test_tril.py,,yes,tril_
+main,triu,triu,benchmark/test_triu.py,,yes,triu
+main,triu_inplace,triu,benchmark/test_triu.py,,yes,triu_
+main,true_divide,true_divide,benchmark/test_true_divide.py,,yes,div_tensor
+main,true_divide_scalar,true_divide,benchmark/test_true_divide.py,,yes,div_scalar
+main,true_divide_inplace_scalar,true_divide,benchmark/test_true_divide.py,,yes,div_scalar_
+main,trunc,trunc_,benchmark/test_trunc_.py,,yes,trunc
+main,trunc_,trunc_,benchmark/test_trunc_.py,,yes,trunc_
+main,trunc_divide,trunc_divide,benchmark/test_trunc_divide.py,,yes,trunc_divide
+main,trunc_divide_inplace,trunc_divide,benchmark/test_trunc_divide.py,,yes,trunc_divide_
+main,unbind,unbind,benchmark/test_unbind.py,,yes,unbind
+main,unbind_copy,unbind_copy,benchmark/test_unbind_copy.py,,yes,unbind_copy
+main,unfold_backward,unfold,benchmark/test_unfold.py,,yes,unfold_backward
+main,unfold_copy,unfold_copy,benchmark/test_unfold_copy.py,,yes,unfold_copy
+main,unfold_view,unfold_view,benchmark/test_unfold_view.py,,yes,unfold
+main,uniform_inplace,uniform,benchmark/test_uniform.py,,yes,uniform_
+main,uniform_,uniform_,benchmark/test_uniform_.py,,yes,uniform_
+main,unique2,unique,benchmark/test_unique.py,,yes,unique2
+main,unique_consecutive,unique_consecutive,benchmark/test_unique_consecutive.py,,yes,unique_consecutive
+main,unique_dim_dim0,unique_dim,benchmark/test_unique_dim.py,,yes,unique_dim
+main,unique_dim_dim1,unique_dim,benchmark/test_unique_dim.py,,yes,unique_dim
+main,unpack_seq,unpack_seq,benchmark/test_unpack_seq.py,,yes,unpack_seq_triton
+main,unpack_seq_fp8,unpack_seq,benchmark/test_unpack_seq.py,,yes,unpack_seq_triton
+main,unsafe_chunk,unsafe_chunk,benchmark/test_unsafe_chunk.py,,yes,unsafe_chunk
+main,unsafe_masked_index,unsafe_masked_index,benchmark/test_unsafe_masked_index.py,,yes,unsafe_masked_index
+main,unsafe_masked_index_put_accumulate,unsafe_masked_index_put_accumulate,benchmark/test_unsafe_masked_index_put_accumulate.py,,yes,unsafe_masked_index_put_accumulate
+main,unsafe_view,unsafe_view,benchmark/test_unsafe_view.py,,yes,unsafe_view
+main,unsqueeze,unsqueeze,benchmark/test_unsqueeze.py,,yes,unsqueeze
+main,unsqueeze_,unsqueeze,benchmark/test_unsqueeze.py,,yes,unsqueeze_
+main,upsample_bicubic2d,upsample_bicubic2d,benchmark/test_upsample_bicubic2d.py,yes,yes,upsample_bicubic2d
+main,upsample_bicubic2d_aa,upsample_bicubic2d_aa,benchmark/test_upsample_bicubic2d_aa.py,,yes,upsample_bicubic2d_aa
+main,upsample_bicubic2d_aa_backward,upsample_bicubic2d_aa,benchmark/test_upsample_bicubic2d_aa.py,,yes,upsample_bicubic2d_aa_backward
+main,upsample_bilinear2d_aa,upsample_bilinear2d_aa,benchmark/test_upsample_bilinear2d_aa.py,,yes,upsample_bilinear2d_aa
+main,upsample_linear1d,upsample_linear1d,benchmark/test_upsample_linear1d.py,,yes,upsample_linear1d
+main,upsample_linear1d_backward,upsample_linear1d,benchmark/test_upsample_linear1d.py,,yes,upsample_linear1d_backward
+main,upsample_nearest1d,upsample_nearest1d,benchmark/test_upsample_nearest1d.py,,yes,upsample_nearest1d
+main,upsample_nearest2d,upsample_nearest2d,benchmark/test_upsample_nearest2d.py,,yes,upsample_nearest2d
+main,upsample_nearest3d,upsample_nearest3d,benchmark/test_upsample_nearest3d.py,,yes,upsample_nearest3d
+main,upsample_nearest_exact1d,upsample_nearest_exact1d,benchmark/test_upsample_nearest_exact1d.py,,yes,upsample_nearest_exact1d
+main,upsample_nearest_exact2d,upsample_nearest_exact2d,benchmark/test_upsample_nearest_exact2d.py,,yes,upsample_nearest_exact2d
+main,upsample_nearest_exact2d_out,upsample_nearest_exact2d,benchmark/test_upsample_nearest_exact2d.py,,yes,upsample_nearest_exact2d
+main,upsample_nearest_exact2d_backward,upsample_nearest_exact2d_backward,benchmark/test_upsample_nearest_exact2d_backward.py,,yes,upsample_nearest_exact2d_backward
+main,upsample_nearest_exact3d,upsample_nearest_exact3d,benchmark/test_upsample_nearest_exact3d.py,,yes,_upsample_nearest_exact3d
+main,upsample_trilinear3d,upsample_trilinear3d,benchmark/test_upsample_trilinear3d.py,,yes,upsample_trilinear3d
+main,var,var,benchmark/test_var.py,,yes,var
+main,var_dim,var,benchmark/test_var.py,,yes,var_dim
+main,var_correction,var,benchmark/test_var.py,,yes,var_correction
+main,var_mean,var_mean,benchmark/test_var_mean.py,,yes,var_mean
+main,vdot,vdot,benchmark/test_vdot.py,,yes,vdot
+main,vector_norm,vector_norm,benchmark/test_vector_norm.py,,yes,vector_norm
+main,view_copy,view_copy,benchmark/test_view_copy.py,,yes,view_copy
+main,vstack,vstack,benchmark/test_vstack.py,,yes,vstack
+main,perf_w8a8_block_fp8_matmul,w8a8_block_fp8_matmul,benchmark/test_w8a8_block_fp8_matmul.py,,yes,w8a8_block_fp8_matmul
+main,weight_int4pack_mm_with_scales_and_zeros,weight_int4pack_mm_with_scales_and_zeros,benchmark/test_weight_int4pack_mm_with_scales_and_zeros.py,,yes,weight_int4pack_mm_with_scales_and_zeros
+main,weight_norm_dim0,weight_norm,benchmark/test_weight_norm.py,,yes,weight_norm
+main,weight_norm_dim_last,weight_norm,benchmark/test_weight_norm.py,,yes,weight_norm
+main,weight_norm_interface,weight_norm_interface,benchmark/test_weight_norm_interface.py,,yes,weight_norm_interface
+main,weight_norm_interface_backward,weight_norm_interface,benchmark/test_weight_norm_interface.py,,yes,weight_norm_interface_backward
+main,where_self,where,benchmark/test_where.py,,yes,where_self
+main,where_self_out,where,benchmark/test_where.py,,yes,where_self_out
+main,xlogy,xlogy,benchmark/test_xlogy.py,,yes,xlogy
+main,xlogy_out,xlogy,benchmark/test_xlogy.py,,yes,xlogy_out
+main,xlogy_tensor_scalar,xlogy,benchmark/test_xlogy.py,,yes,xlogy_tensor_scalar
+main,xlogy_tensor_scalar_out,xlogy,benchmark/test_xlogy.py,,yes,xlogy_tensor_scalar_out
+main,xlogy_scalar_tensor,xlogy,benchmark/test_xlogy.py,,yes,xlogy_scalar_tensor
+main,xlogy_scalar_tensor_out,xlogy,benchmark/test_xlogy.py,,yes,xlogy_scalar_tensor_out
+main,xlogy_,xlogy_,benchmark/test_xlogy_.py,,yes,xlogy_
+main,xlogy_scalar_other_,xlogy_,benchmark/test_xlogy_.py,,yes,xlogy_scalar_other_
+main,xor__,xor__,benchmark/test_xor__.py,,yes,xor__
+main,zero,zero,benchmark/test_zero.py,,yes,zero
+main,zero_inplace,zero,benchmark/test_zero.py,,yes,zero_
+main,zero_out,zero,benchmark/test_zero.py,,yes,zero_out
+main,zeros,zeros,benchmark/test_zeros.py,,yes,zeros
+main,zeros_like,zeros_like,benchmark/test_zeros_like.py,,yes,zeros_like
diff --git a/ops/inventory_vllm.csv b/ops/inventory_vllm.csv
new file mode 100644
index 0000000..8a7ed81
--- /dev/null
+++ b/ops/inventory_vllm.csv
@@ -0,0 +1,109 @@
+repo,op,op_file,file,marked_skip,collected,op_names
+vllm,perf_chunk_gated_delta_rule_fwd,FLA/test_chunk_gated_delta_rule_fwd,benchmark/test_FLA/test_chunk_gated_delta_rule_fwd.py,,yes,chunk_gated_delta_rule_fwd
+vllm,perf_chunk_gated_delta_rule,FLA/test_chunk_gated_delta_rule_perf,benchmark/test_FLA/test_chunk_gated_delta_rule_perf.py,,yes,chunk_gated_delta_rule
+vllm,chunk_gdn2,FLA/test_chunk_gdn2,benchmark/test_FLA/test_chunk_gdn2.py,,yes,chunk_gdn2
+vllm,perf_chunk_gla,FLA/test_chunk_gla_perf,benchmark/test_FLA/test_chunk_gla_perf.py,,yes,
+vllm,chunk_kda,FLA/test_chunk_kda,benchmark/test_FLA/test_chunk_kda.py,,yes,chunk_kda
+vllm,perf_fused_recurrent_gated_delta_rule,FLA/test_fused_recurrent_gated_delta_rule_perf,benchmark/test_FLA/test_fused_recurrent_gated_delta_rule_perf.py,,yes,fused_recurrent_gated_delta_rule
+vllm,act_quant_perf,act_quant,benchmark/test_act_quant.py,,yes,act_quant_triton
+vllm,add_rms_norm,add_rms_norm,benchmark/test_add_rms_norm.py,,yes,add_rms_norm
+vllm,apply_repetition_penalties,apply_repetition_penalties,benchmark/test_apply_repetition_penalties.py,,yes,apply_repetition_penalties
+vllm,apply_rotary_pos_emb,apply_rotary_pos_emb,benchmark/test_apply_rotary_pos_emb.py,,yes,apply_rotary_pos_emb
+vllm,beam_search_score,beam_search_score,benchmark/test_beam_search_score.py,,yes,beam_search_score
+vllm,beam_search_score_,beam_search_score,benchmark/test_beam_search_score.py,,yes,beam_search_score_
+vllm,bincount,bincount,benchmark/test_bincount.py,,yes,bincount
+vllm,bincount_weighted,bincount,benchmark/test_bincount.py,,yes,bincount_weighted
+vllm,blas_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,
+vllm,perf_w8a8_block_fp8_matmul,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,w8a8_block_fp8_matmul
+vllm,perf_w8a8_block_fp8_matmul_deepgemm,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,w8a8_block_fp8_matmul_deepgemm
+vllm,perf_sparse_attention,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,sparse_attention
+vllm,mv_and_outer_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,
+vllm,addmv_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,
+vllm,vdot_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,vdot
+vllm,addr_benchmark,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,addr
+vllm,perf_router_gemm,blas_perf_parallel,benchmark/test_blas_perf_parallel.py,,yes,router_gemm
+vllm,concat_and_cache_mla,concat_and_cache_mla,benchmark/test_concat_and_cache_mla.py,,yes,concat_and_cache_mla
+vllm,cp_gather_indexer_k_quant_cache_benchmark,cp_gather_indexer_k_quant_cache,benchmark/test_cp_gather_indexer_k_quant_cache.py,,yes,
+vllm,cross_entropy_loss,cross_entropy_loss,benchmark/test_cross_entropy_loss.py,,yes,cross_entropy_loss
+vllm,cutlass_scaled_mm_benchmark,cutlass_scaled_mm,benchmark/test_cutlass_scaled_mm.py,,yes,
+vllm,combine_topk_swa_indices_benchmark,deepseek_v4_attention_combine_topk_swa_indices,benchmark/test_deepseek_v4_attention_combine_topk_swa_indices.py,,yes,
+vllm,compute_global_topk_indices_and_lens_benchmark,deepseek_v4_attention_compute_global_topk_indices_and_lens,benchmark/test_deepseek_v4_attention_compute_global_topk_indices_and_lens.py,,yes,
+vllm,dequantize_and_gather_k_cache_benchmark,deepseek_v4_attention_dequantize_and_gather_k_cache,benchmark/test_deepseek_v4_attention_dequantize_and_gather_k_cache.py,,yes,
+vllm,fused_q_kv_rmsnorm_benchmark,deepseek_v4_attention_fused_q_kv_rmsnorm,benchmark/test_deepseek_v4_attention_fused_q_kv_rmsnorm.py,,yes,
+vllm,dgeglu,dgeglu,benchmark/test_dgeglu.py,,yes,dgeglu
+vllm,dreglu,dreglu,benchmark/test_dreglu.py,,yes,dreglu
+vllm,dswiglu,dswiglu,benchmark/test_dswiglu.py,,yes,dswiglu
+vllm,flash_attention_forward,flash_attention_forward,benchmark/test_flash_attention_forward.py,,yes,flash_attention_forward
+vllm,flash_attn_varlen_func,flash_attn_varlen_func,benchmark/test_flash_attn_varlen_func.py,,yes,flash_attn_varlen_func
+vllm,flash_attn_varlen_opt_func,flash_attn_varlen_opt_init_func,benchmark/test_flash_attn_varlen_opt_init_func.py,,yes,flash_attn_varlen_func
+vllm,flash_mla,flash_mla,benchmark/test_flash_mla.py,,yes,flash_mla
+vllm,flash_mla_sparse_fwd,flash_mla_sparse_fwd,benchmark/test_flash_mla_sparse_fwd.py,,yes,
+vllm,flash_mla_with_kvcache,flash_mla_with_kvcache,benchmark/test_flash_mla_with_kvcache.py,,yes,
+vllm,fp8_fp4_mqa_logits,fp8_fp4_mqa_logits,benchmark/test_fp8_fp4_mqa_logits.py,,yes,fp8_fp4_mqa_logits
+vllm,fp8_fp4_paged_mqa_logits,fp8_fp4_paged_mqa_logits,benchmark/test_fp8_fp4_paged_mqa_logits.py,,yes,fp8_fp4_paged_mqa_logits
+vllm,fused_add_rms_norm,fused_add_rms_norm,benchmark/test_fused_add_rms_norm.py,,yes,fused_add_rms_norm
+vllm,fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert,fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert,benchmark/test_fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert.py,,yes,
+vllm,fused_indexer_q_rope_quant,fused_indexer_q_rope_quant,benchmark/test_fused_indexer_q_rope_quant.py,,yes,fused_indexer_q_rope_quant
+vllm,fused_inv_rope_fp8_quant,fused_inv_rope_fp8_quant,benchmark/test_fused_inv_rope_fp8_quant.py,,yes,fused_inv_rope_fp8_quant
+vllm,fused_marlin_moe,fused_marlin_moe,benchmark/test_fused_marlin_moe.py,,yes,fused_marlin_moe
+vllm,fused_moe_impl_gems_vs_vllm,fused_moe,benchmark/test_fused_moe.py,,yes,fused_experts_impl
+vllm,fused_moe_fp8,fused_moe_fp8,benchmark/test_fused_moe_fp8.py,,yes,fused_experts_impl
+vllm,fused_moe_fp8_blockwise,fused_moe_fp8_blockwise,benchmark/test_fused_moe_fp8_blockwise.py,,yes,fused_experts_impl
+vllm,fused_moe_int4_w4a16,fused_moe_int4_w4a16,benchmark/test_fused_moe_int4_w4a16.py,,yes,fused_experts_impl
+vllm,fused_experts_impl_int8,fused_moe_int8,benchmark/test_fused_moe_int8.py,,yes,fused_experts_impl
+vllm,fused_experts_impl_int8_w8a16,fused_moe_int8_w8a16,benchmark/test_fused_moe_int8_w8a16.py,,yes,fused_moe_int8_w8a16_gems_vs_bf16_deq
+vllm,fused_moe_w8a16_mxq,fused_moe_w8a16,benchmark/test_fused_moe_w8a16.py,,yes,fused_moe_w8a16_mxq_gems_vs_bf16_deq
+vllm,fused_moe_w8a16_mxq_gems_vs_vllm,fused_moe_w8a16,benchmark/test_fused_moe_w8a16.py,,yes,fused_moe_w8a16_mxq_gems_vs_vllm
+vllm,geglu,geglu,benchmark/test_geglu.py,,yes,geglu
+vllm,gelu_and_mul,gelu_and_mul,benchmark/test_gelu_and_mul.py,,yes,gelu_and_mul
+vllm,grouped_topk_no_renorm,grouped_topk,benchmark/test_grouped_topk.py,,yes,grouped_topk
+vllm,grouped_topk_score_0,grouped_topk,benchmark/test_grouped_topk.py,,yes,grouped_topk
+vllm,grouped_topk_score_1,grouped_topk,benchmark/test_grouped_topk.py,,yes,grouped_topk
+vllm,indexer_k_quant_and_cache_benchmark,indexer_k_quant_and_cache,benchmark/test_indexer_k_quant_and_cache.py,,yes,
+vllm,inplace_fused_experts_gems_vs_vllm,inplace_fused_experts,benchmark/test_inplace_fused_experts.py,,yes,inplace_fused_experts
+vllm,instance_norm,instance_norm,benchmark/test_instance_norm.py,,yes,instance_norm
+vllm,mhc_post,mhc,benchmark/test_mhc.py,,yes,mhc_post
+vllm,mhc_pre,mhc,benchmark/test_mhc.py,,yes,mhc_pre
+vllm,hc_split_sinkhorn_forward,mhc,benchmark/test_mhc.py,,yes,hc_split_sinkhorn_forward
+vllm,mhc_bwd,mhc,benchmark/test_mhc.py,,yes,mhc_bwd
+vllm,hc_head_fused_kernel,mhc,benchmark/test_mhc.py,,yes,hc_head_fused_kernel
+vllm,moe_align_block_size_triton,moe_align_block_size_triton,benchmark/test_moe_align_block_size_triton.py,,yes,moe_align_block_size_triton
+vllm,moe_sum,moe_sum,benchmark/test_moe_sum.py,,yes,moe_sum
+vllm,mrope_gems_vs_torch,mrope,benchmark/test_mrope.py,,yes,mrope
+vllm,outer,outer,benchmark/test_outer.py,,yes,outer
+vllm,outplace_fused_experts_gems_vs_vllm,outplace_fused_experts,benchmark/test_outplace_fused_experts.py,,yes,outplace_fused_experts
+vllm,pack_seq,pack_seq,benchmark/test_pack_seq.py,,yes,pack_seq_triton
+vllm,pack_seq_fp8,pack_seq,benchmark/test_pack_seq.py,,yes,pack_seq_triton
+vllm,perf_parallel_nsa,parallel_nsa,benchmark/test_parallel_nsa.py,,yes,parallel_nsa
+vllm,perf_parallel_nsa_compression,parallel_nsa_compression,benchmark/test_parallel_nsa_compression.py,,yes,parallel_nsa_compression
+vllm,per_token_group_quant_fp8,per_token_group_quant_fp8,benchmark/test_per_token_group_quant_fp8.py,,yes,per_token_group_quant_fp8
+vllm,persistent_topk,persistent_topk,benchmark/test_persistent_topk.py,,yes,persistent_topk
+vllm,reglu,reglu,benchmark/test_reglu.py,,yes,reglu
+vllm,reshape_and_cache,reshape_and_cache,benchmark/test_reshape_and_cache.py,,yes,reshape_and_cache
+vllm,reshape_and_cache_flash,reshape_and_cache_flash,benchmark/test_reshape_and_cache_flash.py,,yes,reshape_and_cache_flash
+vllm,perf_router_gemm,router_gemm,benchmark/test_router_gemm.py,,yes,router_gemm
+vllm,rwkv_ka_fusion,rwkv_ka_fusion,benchmark/test_rwkv_ka_fusion.py,,yes,rwkv_ka_fusion
+vllm,rwkv_mm_sparsity,rwkv_mm_sparsity,benchmark/test_rwkv_mm_sparsity.py,,yes,rwkv_mm_sparsity
+vllm,dynamic_scaled_int8_quant,scaled_int8_quant,benchmark/test_scaled_int8_quant.py,,yes,dynamic_scaled_int8_quant
+vllm,static_scaled_int8_quant,scaled_int8_quant,benchmark/test_scaled_int8_quant.py,,yes,static_scaled_int8_quant
+vllm,silu_and_mul,silu_and_mul,benchmark/test_silu_and_mul.py,,yes,silu_and_mul
+vllm,silu_and_mul_out,silu_and_mul,benchmark/test_silu_and_mul.py,,yes,silu_and_mul_out
+vllm,silu_and_mul_with_clamp,silu_and_mul_with_clamp,benchmark/test_silu_and_mul_with_clamp.py,,yes,silu_and_mul_with_clamp
+vllm,silu_and_mul_with_clamp_out,silu_and_mul_with_clamp,benchmark/test_silu_and_mul_with_clamp.py,,yes,silu_and_mul_with_clamp_out
+vllm,skip_layernorm,skip_layer_norm,benchmark/test_skip_layer_norm.py,,yes,skip_layer_norm
+vllm,sparse_attn_triton,sparse_attention,benchmark/test_sparse_attention.py,yes,yes,sparse_attention
+vllm,sparse_mla_fwd_interface,sparse_mla_fwd_interface,benchmark/test_sparse_mla_fwd_interface.py,,yes,sparse_mla_fwd_interface
+vllm,stage_deepseek_v4_mega_moe_inputs_benchmark,stage_deepseek_v4_mega_moe_inputs,benchmark/test_stage_deepseek_v4_mega_moe_inputs.py,,yes,
+vllm,swiglu,swiglu,benchmark/test_swiglu.py,,yes,swiglu
+vllm,top_k_per_row_decode,top_k_per_row_decode,benchmark/test_top_k_per_row_decode.py,,yes,top_k_per_row_decode
+vllm,top_k_per_row_prefill,top_k_per_row_prefill,benchmark/test_top_k_per_row_prefill.py,,yes,top_k_per_row_prefill
+vllm,topk_softmax,topk_softmax,benchmark/test_topk_softmax.py,,yes,topk_softmax
+vllm,topk_softplus_sqrt,topk_softplus_sqrt,benchmark/test_topk_softplus_sqrt.py,,yes,topk_softplus_sqrt
+vllm,triton_scaled_mm_benchmark,triton_scaled_mm,benchmark/test_triton_scaled_mm.py,,yes,triton_scaled_mm
+vllm,triton_unified_attention_perf,triton_unified_attention_perf,benchmark/test_triton_unified_attention_perf.py,,yes,
+vllm,unpack_seq,unpack_seq,benchmark/test_unpack_seq.py,,yes,unpack_seq_triton
+vllm,unpack_seq_fp8,unpack_seq,benchmark/test_unpack_seq.py,,yes,unpack_seq_triton
+vllm,weight_norm_dim0,weight_norm,benchmark/test_weight_norm.py,,yes,weight_norm
+vllm,weight_norm_dim_last,weight_norm,benchmark/test_weight_norm.py,,yes,weight_norm
+vllm,weight_norm_interface,weight_norm_interface,benchmark/test_weight_norm_interface.py,,yes,weight_norm_interface
+vllm,weight_norm_interface_backward,weight_norm_interface,benchmark/test_weight_norm_interface.py,,yes,weight_norm_interface_backward
diff --git a/ops/shapes_multi.yaml b/ops/shapes_multi.yaml
new file mode 100644
index 0000000..c4e9daa
--- /dev/null
+++ b/ops/shapes_multi.yaml
@@ -0,0 +1,9639 @@
+# 由 gen_inventory.py 从 shapes_multi.yaml 迁移：保留新上游仍存在的 op_name 键 555/601，并合并上游 core_shapes.yaml 缺失键（main:+43 vllm:+3）。
+# 顶层键 = FlagGems op_name 或 Benchmark 类名（上游回退链需要）；
+# 两者都无键的算子使用其基类 DEFAULT_SHAPES。
+AttentionBenchmark:
+  shapes:
+  - [4, 32, 1024, 64]
+  - [4, 32, 1024, 128]
+  - [4, 32, 2048, 128]
+  - [4, 32, 4096, 128]
+  - [4, 32, 8192, 128]
+Benchmark:
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+BlasBenchmark:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+ChunkGatedDeltaRuleBenchmark:
+  shape_desc: B, T, Hg, H, K, V
+  shapes:
+  - [1, 64, 4, 8, 64, 64]
+  - [1, 128, 4, 8, 64, 64]
+  - [1, 256, 4, 8, 64, 64]
+  - [1, 512, 4, 8, 64, 64]
+  - [2, 64, 4, 8, 64, 64]
+  - [2, 128, 4, 8, 64, 64]
+  - [2, 256, 4, 8, 64, 64]
+  - [2, 512, 4, 8, 64, 64]
+ConcatAndCacheMLABenchmark:
+  shapes:
+  - [16, 8, 84, 32, 4]
+  - [32, 4, 21, 16, 16]
+  - [64, 8, 21, 8, 4]
+  - [512, 64, 42, 16, 8]
+  - [1024, 64, 84, 16, 8]
+ConcatBenchmark:
+  shape_desc: ((B), M, N) * 3
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [512, 512]
+  - [512, 1024]
+  - [512, 2048]
+Conv1DBenchmark:
+  shape_desc: N, C, L, Co, K, stride, padding, groups
+  shapes:
+  - [32, 64, 256, 64, 3, 1, 1, 1]
+  - [64, 32, 512, 32, 5, 2, 2, 1]
+  - [16, 16, 1024, 32, 7, 1, 3, 1]
+  - [32, 32, 4096, 64, 9, 2, 4, 2]
+Conv2DBenchmark:
+  shape_desc: N, C_in, H, W, C_out, K_h, K_w, stride, padding, groups
+  shapes:
+  - [32, 64, 128, 128, 32, 3, 3, 1, 1, 1]
+  - [104, 16, 32, 32, 32, 4, 4, 1, 0, 1]
+  - [64, 32, 18, 18, 32, 5, 5, 2, 1, 1]
+  - [64, 32, 210, 210, 64, 5, 5, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 1, 1, 2]
+Conv3DBenchmark:
+  shape_desc: N, C_in, D, H, W, C_out, K_d, K_h, K_w, stride, padding, groups
+  shapes:
+  - [32, 64, 64, 64, 64, 32, 3, 3, 3, 1, 1, 1]
+  - [32, 64, 64, 64, 64, 32, 3, 3, 3, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 24, 3, 3, 3, 1, 1, 2]
+  - [64, 32, 18, 18, 18, 32, 5, 5, 5, 1, 2, 1]
+  - [16, 32, 12, 12, 12, 24, 3, 3, 3, 1, 2, 1]
+ConvDepthwise2DBenchmark:
+  shape_desc: N, C, H, W, K_h, K_w, stride, padding, dilation
+  shapes:
+  - [32, 32, 64, 64, 3, 3, 1, 1, 1]
+  - [16, 64, 128, 128, 3, 3, 2, 1, 1]
+  - [8, 128, 56, 56, 5, 5, 1, 2, 1]
+  - [4, 256, 28, 28, 3, 3, 1, 1, 2]
+  - [2, 512, 14, 14, 3, 3, 1, 1, 1]
+EmbeddingBenchmark:
+  shapes:
+  - [4, 4]
+  - [16, 16]
+  - [128, 128]
+  - [256, 256]
+  - [1024, 1024]
+FlashMLABenchmark:
+  shapes:
+  - [1024]
+  - [2048]
+  - [4096]
+  - [8192]
+  - [16384]
+FractionalMaxPool2dBenchmark:
+  shape_desc: N, C, H, W
+  shapes:
+  - [4, 3, 224, 224]
+  - [16, 64, 56, 56]
+  - [32, 128, 28, 28]
+  - [64, 256, 14, 14]
+  - [128, 512, 7, 7]
+FusedRecurrentGatedDeltaRuleBenchmark:
+  shape_desc: (T,) many lengths
+  shapes:
+  - [1]
+  - [64]
+  - [184]
+  - [256]
+  - [512]
+GenericBenchmark2DOnly:
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+GenericBenchmark4DOnly:
+  shapes:
+  - [64, 64, 4, 8]
+  - [256, 256, 4, 8]
+  - [1024, 1024, 4, 8]
+  - [4096, 4096, 4, 8]
+  - [1024, 65536, 4, 8]
+GenericBenchmarkExcluse1D:
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+GenericBenchmarkExcluse3D:
+  shapes:
+  - [1048576]
+  - [1073741824]
+  - [1024, 512]
+  - [1024, 1024]
+  - [4096, 4096]
+IndexPutAccFalseBenchmark:
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [2, 8]
+    - [8]
+    - false
+  - - [1024, 1024]
+    - - [4, 64]
+    - [1024]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+    - [512]
+    - false
+IndexPutAccTrueBenchmark:
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [8]
+    - [8]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [64]
+    - [64]
+    - false
+  - - [512, 512, 512]
+    - - [128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [2, 128]
+      - [2, 128]
+    - [2, 128]
+    - false
+IndexReduceBenchmark:
+  shape_desc: (B), M, N
+  shapes:
+  - [1048576]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+KronBenchmark:
+  shapes:
+  - [16, 16]
+  - [64, 64]
+  - [128, 128]
+  - [256, 256]
+  - [4, 8, 16, 32]
+MoeAlignBlockSizeBenchmark:
+  shapes:
+  - [16, 8]
+  - [4, 32]
+  - [32, 16]
+  - [8, 64]
+  - [128, 4]
+MvAndOuterBenchmark:
+  shapes:
+  - [384, 384]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [8192, 8192]
+  - [10240, 10240]
+NormBenchmark:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+PerTokenGroupQuantFp8Benchmark:
+  shapes:
+  - [7, 512, 512]
+  - [7, 4096, 256]
+  - [83, 512, 64]
+  - [2048, 4096, 256]
+  - [2048, 13824, 512]
+RWKVBenchmark:
+  shapes:
+  - [64]
+  - [256]
+  - [1024]
+  - [4096]
+  - [8192]
+  - [16384]
+  - [32768]
+RWKVSparsityBenchmark:
+  shapes:
+  - [1]
+ReshapeAndCacheBenchmark:
+  shapes:
+  - [21, 4, 64, 8, 512]
+  - [42, 8, 64, 8, 1024]
+  - [42, 8, 80, 16, 10000]
+  - [42, 8, 120, 32, 10000]
+  - [42, 8, 256, 8, 1024]
+ReshapeAndCacheFlashBenchmark:
+  shapes:
+  - [21, 4, 64, 8, 512]
+  - [21, 4, 64, 8, 1024]
+  - [21, 4, 80, 16, 10000]
+  - [42, 8, 64, 8, 1024]
+  - [42, 8, 80, 16, 10000]
+RopeBenchmark:
+  shapes:
+  - [128]
+  - [192]
+  - [256]
+  - [384]
+  - [512]
+ScaledDotProductFusedAttentionOverrideableBenchmark:
+  shape_desc: B, H, S, D
+  shapes:
+  - [2, 8, 64, 64]
+  - [4, 8, 128, 64]
+  - [2, 8, 128, 128]
+  - [1, 16, 256, 64]
+TensorRepeatBenchmark:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [512, 512]
+  - [512, 1024]
+  - [512, 2048]
+TensorSelectBenchmark:
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+UnaryPointwiseBenchmark:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+UnaryReductionBenchmark:
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+UpsampleBenchmark:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+abs:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+abs_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+absolute:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+acos:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+act_quant_triton:
+  shape_desc: M, N
+  shapes:
+  - [1, 128]
+  - [1, 896]
+  - [1, 2048]
+  - [1, 8192]
+  - [40, 128]
+  - [40, 896]
+  - [40, 2048]
+  - [40, 8192]
+  - [164, 128]
+  - [164, 896]
+  - [164, 2048]
+  - [164, 8192]
+  - [512, 128]
+  - [512, 896]
+  - [512, 2048]
+  - [512, 8192]
+  - [3454, 128]
+  - [3454, 896]
+  - [3454, 2048]
+  - [3454, 8192]
+  - [12027, 128]
+  - [12027, 896]
+  - [12027, 2048]
+  - [12027, 8192]
+  - [38594, 128]
+  - [38594, 896]
+  - [38594, 2048]
+  - [38594, 8192]
+add_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+add_rms_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+addcdiv:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+addcdiv_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+addcmul:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+addcmul_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+addmm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+addmm_dtype:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+addmm_dtype_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+addmm_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+addmv:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+addr:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+alias_copy:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+alias_copy_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+all:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+all_dim:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+allclose:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+amax:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+aminmax:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+angle:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+any:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+any_dim:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+any_dims:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+apply_repetition_penalties:
+  shape_desc: M, N
+  shapes:
+  - [1, 1024]
+  - [1, 4096]
+  - [1, 8192]
+  - [8, 4096]
+  - [16, 4096]
+  - [32, 1024]
+  - [8, 8192]
+  - [64, 32000]
+apply_rotary_pos_emb:
+  shape_desc: M, N
+  shapes:
+  - [128]
+  - [192]
+  - [256]
+  - [384]
+  - [512]
+arange:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+arange_start:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+arange_start_step:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+arcsinh:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+arcsinh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+arcsinh_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+arctanh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+argmax:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+argmin:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+argsort:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 512]
+  - [16, 131072]
+  - [8, 262144]
+as_strided_copy:
+  shape_desc: input shape, size, stride, storage_offset
+  shapes:
+  - - [64, 64]
+    - [64, 64]
+    - [64, 1]
+    - 0
+  - - [256, 256]
+    - [128, 128]
+    - [1, 256]
+    - 0
+  - - [1024, 1024]
+    - [512, 512]
+    - [1024, 1]
+    - 0
+  - - [64, 128, 64]
+    - [32, 64, 32]
+    - [8192, 64, 2]
+    - 0
+  - - [1048576]
+    - [524288]
+    - [2]
+    - 0
+as_strided_copy_out:
+  shape_desc: input shape, size, stride, storage_offset
+  shapes:
+  - - [64, 64]
+    - [64, 64]
+    - [64, 1]
+    - 0
+  - - [256, 256]
+    - [128, 128]
+    - [1, 256]
+    - 0
+  - - [1024, 1024]
+    - [512, 512]
+    - [1024, 1]
+    - 0
+  - - [64, 128, 64]
+    - [32, 64, 32]
+    - [8192, 64, 2]
+    - 0
+  - - [1048576]
+    - [524288]
+    - [2]
+    - 0
+asinh:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+asinh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+assert_async:
+  shape_desc: M, N
+  shapes:
+  - []
+  - [1]
+  - [1, 1]
+  - [1, 1, 1]
+atan:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+atan2:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+atan2_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+atan_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+atanh:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+avg_pool2d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [4, 3, 224, 224]
+  - [16, 64, 56, 56]
+  - [32, 128, 28, 28]
+  - [64, 256, 14, 14]
+  - [128, 512, 7, 7]
+avg_pool2d_backward:
+  shape_desc: N, C, H, W
+  shapes:
+  - [4, 3, 224, 224]
+  - [16, 64, 56, 56]
+  - [32, 128, 28, 28]
+  - [64, 256, 14, 14]
+  - [128, 512, 7, 7]
+avg_pool3d:
+  shape_desc: N, C, D, H, W
+  shapes:
+  - [4, 3, 16, 56, 56]
+  - [8, 64, 8, 28, 28]
+  - [16, 128, 4, 14, 14]
+  - [32, 256, 4, 7, 7]
+baddbmm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+batch_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+  - [16, 16, 64]
+  - [16, 16, 1024]
+  - [16, 16, 4098]
+  - [1, 8, 4, 4]
+  - [16, 8, 128, 128]
+batch_norm_backward:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+  - [16, 16, 64]
+  - [16, 16, 1024]
+  - [16, 16, 4098]
+  - [1, 8, 4, 4]
+  - [16, 8, 128, 128]
+bernoulli_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bincount:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bincount_weighted:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_and_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_and_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_and_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_and_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+bitwise_and_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+bitwise_left_shift:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+bitwise_not:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+bitwise_not_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+bitwise_or_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_or_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_or_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+bitwise_or_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+bitwise_or_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+bitwise_right_shift:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+bmm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+bmm_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+cat_out:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [64, 64, 1]
+  - [64, 64, 16]
+cauchy:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+cauchy_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+cdist_backward:
+  shape_desc: batch, n1, dim
+  shapes:
+  - [2, 16, 32]
+  - [4, 32, 64]
+  - [8, 64, 128]
+ceil:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+ceil_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+ceil_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+celu:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+celu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+chunk_gated_delta_rule_fwd:
+  shape_desc: B, T, H, K, V
+  shapes:
+  - [2, 16384, 16, 128, 128]
+  - [4, 2048, 16, 128, 128]
+  - [4, 4096, 64, 128, 128]
+chunk_gdn2:
+  shapes:
+  - [2, 512, 8, 64, 64]
+  - [4, 1024, 8, 64, 64]
+  - [1, 2048, 8, 64, 64]
+  - [1, 4096, 16, 64, 64]
+  - [1, 8192, 96, 128, 128]
+  - [2, 2048, 16, 256, 512]
+  - [2, 16384, 16, 128, 128]
+  - [4, 1024, 8, 256, 512]
+  - [4, 2048, 16, 128, 128]
+  - [4, 4096, 64, 128, 128]
+  - [8, 1024, 8, 64, 64]
+  - [8, 2048, 32, 256, 256]
+clamp:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_max:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_max_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_min:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_min_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clamp_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clip:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+clip_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+col2im:
+  shape_desc: M, N
+  shapes:
+  - - 2
+    - 3
+    - [2, 2]
+    - [4, 5]
+    - [1, 1]
+    - [0, 0]
+    - [1, 1]
+  - - 2
+    - 16
+    - [3, 3]
+    - [16, 16]
+    - [1, 1]
+    - [0, 0]
+    - [1, 1]
+  - - 4
+    - 32
+    - [3, 3]
+    - [32, 32]
+    - [1, 1]
+    - [1, 1]
+    - [1, 1]
+  - - 2
+    - 64
+    - [3, 3]
+    - [64, 64]
+    - [2, 2]
+    - [1, 1]
+    - [1, 1]
+  - - 1
+    - 128
+    - [3, 3]
+    - [128, 128]
+    - [1, 1]
+    - [1, 1]
+    - [1, 1]
+  - - 2
+    - 64
+    - [5, 5]
+    - [64, 64]
+    - [2, 2]
+    - [2, 2]
+    - [1, 1]
+  - - 4
+    - 32
+    - [3, 3]
+    - [128, 128]
+    - [2, 2]
+    - [1, 1]
+    - [2, 2]
+combine_topk_swa_indices:
+  shape_desc: M, N
+  shapes:
+  - - [3, 2]
+    - [6, 4]
+    - [4, 3]
+    - 4
+    - 4
+    - 2
+    - 20
+    - 8
+  - - [128]
+    - [512]
+    - [256]
+    - 32
+    - 128
+    - 4
+    - 42240
+    - 40960
+  - - [512, 256]
+    - [2048, 1024]
+    - [1024, 512]
+    - 64
+    - 256
+    - 4
+    - 45056
+    - 40960
+  - - [4096]
+    - [4096]
+    - [4096]
+    - 128
+    - 256
+    - 4
+    - 45056
+    - 40960
+  - - [1024, 1024]
+    - [8192, 4096]
+    - [2048, 1024]
+    - 128
+    - 256
+    - 4
+    - 45056
+    - 40960
+  - - [128]
+    - [4096]
+    - [512]
+    - 32
+    - 256
+    - 128
+    - 5632
+    - 1280
+  - - [4096]
+    - [4096]
+    - [4096]
+    - 128
+    - 256
+    - 128
+    - 8448
+    - 1280
+compute_global_topk_indices_and_lens:
+  shape_desc: M, N
+  shapes:
+  - [5, 4, 2, 4, 64]
+  - [128, 32, 1, 64, 64]
+  - [512, 64, 2, 128, 64]
+  - [4096, 128, 1, 640, 64]
+  - [4096, 128, 4, 640, 64]
+  - [8192, 128, 8, 1280, 64]
+concat_and_cache_mla:
+  shape_desc: M, N
+  shapes:
+  - [16, 8, 84, 32, 4]
+  - [32, 4, 21, 16, 16]
+  - [64, 8, 21, 8, 4]
+  - [512, 64, 42, 16, 8]
+  - [1024, 64, 84, 16, 8]
+concatenate:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+conj_physical:
+  shape_desc: M, N
+  shapes:
+  - [256]
+  - [2048, 2048]
+  - [128, 512, 256]
+  - [32, 64]
+  - [512, 1024]
+  - [2, 3, 4]
+constant_pad_nd:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+contiguous:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+conv1d:
+  shape_desc: N, C, L, Co, K, stride, padding, groups
+  shapes:
+  - [32, 64, 256, 64, 3, 1, 1, 1]
+  - [64, 32, 512, 32, 5, 2, 2, 1]
+  - [16, 16, 1024, 32, 7, 1, 3, 1]
+  - [32, 32, 4096, 64, 9, 2, 4, 2]
+  - [32, 64, 512, 64, 3, 1, 0, 1]
+  - [64, 48, 1024, 128, 5, 2, 2, 1]
+  - [16, 24, 2048, 96, 7, 1, 3, 2]
+  - [8, 8, 8192, 16, 11, 4, 5, 1]
+  - [4, 4, 16384, 4, 15, 2, 7, 1]
+  - [32, 64, 512, 64, 3, 1, valid, 1]
+  - [64, 48, 1024, 128, 5, 2, valid, 1]
+  - [16, 24, 2048, 96, 7, 1, same, 2]
+  - [8, 8, 8192, 16, 11, 1, same, 1]
+conv1d_padding:
+  shape_desc: batch, input_c, input_l, out_c, kernel, stride, padding, groups
+  shapes:
+  - [32, 64, 512, 64, 3, 1, 0, 1]
+  - [32, 64, 512, 64, 3, 1, 1, 1]
+  - [32, 64, 512, 64, 3, 1, 2, 1]
+  - [64, 48, 1024, 128, 5, 2, 0, 1]
+  - [64, 48, 1024, 128, 5, 2, 2, 1]
+  - [64, 48, 1024, 128, 5, 2, 4, 1]
+  - [16, 24, 2048, 96, 7, 1, 0, 2]
+  - [16, 24, 2048, 96, 7, 1, 3, 2]
+  - [16, 24, 2048, 96, 7, 1, 6, 2]
+  - [32, 64, 512, 64, 3, 1, valid, 1]
+  - [64, 48, 1024, 128, 5, 2, valid, 1]
+  - [16, 24, 2048, 96, 7, 1, same, 2]
+  - [8, 8, 8192, 16, 11, 1, same, 1]
+conv2d:
+  shape_desc: N, C_in, H, W, C_out, K_h, K_w, stride, padding, groups
+  shapes:
+  - [32, 64, 128, 128, 32, 3, 3, 1, 1, 1]
+  - [104, 16, 32, 32, 32, 4, 4, 1, 0, 1]
+  - [64, 32, 18, 18, 32, 5, 5, 2, 1, 1]
+  - [64, 32, 210, 210, 64, 5, 5, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 1, 1, 2]
+  - [32, 64, 128, 128, 32, 3, 3, 1, 2, 1]
+  - [32, 64, 210, 210, 16, 5, 5, 2, 1, 1]
+  - [16, 32, 12, 12, 24, 3, 3, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 2, 2, 2]
+  - [16, 32, 24, 24, 24, 3, 3, 1, 2, 2]
+  - [16, 32, 12, 12, 24, 3, 3, 2, valid, 1]
+  - [32, 64, 128, 128, 32, 3, 3, 1, valid, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 1, same, 2]
+  - [32, 64, 210, 210, 16, 5, 5, 1, same, 1]
+conv2d_padding:
+  shape_desc: batch, input_c, input_h, input_w, out_c, kernel_h, kernel_w, stride,
+    padding, groups
+  shapes:
+  - [16, 32, 12, 12, 24, 3, 3, 1, valid, 1]
+  - [32, 64, 128, 128, 32, 3, 3, 1, valid, 1]
+  - [32, 64, 210, 210, 16, 5, 5, 1, valid, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 1, same, 1]
+  - [32, 64, 128, 128, 32, 3, 3, 1, same, 1]
+  - [32, 64, 210, 210, 16, 5, 5, 1, same, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 1, same, 2]
+conv3d:
+  shape_desc: N, C_in, D, H, W, C_out, K_d, K_h, K_w, stride, padding, groups
+  shapes:
+  - [32, 64, 64, 64, 64, 32, 3, 3, 3, 1, 1, 1]
+  - [32, 64, 64, 64, 64, 32, 3, 3, 3, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 24, 3, 3, 3, 1, 1, 2]
+  - [64, 32, 18, 18, 18, 32, 5, 5, 5, 1, 2, 1]
+  - [16, 32, 12, 12, 12, 24, 3, 3, 3, 1, 2, 1]
+conv3d_padding:
+  shape_desc: batch, input_c, input_d, input_h, input_w, out_c, kernel_d, kernel_h,
+    kernel_w, stride, padding, groups
+  shapes:
+  - [2, 3, 9, 9, 9, 3, 3, 3, 3, 1, valid, 1]
+  - [9, 16, 4, 4, 4, 128, 2, 2, 2, 1, valid, 4]
+  - [32, 8, 8, 8, 8, 32, 3, 3, 3, 1, valid, 1]
+  - [2, 3, 9, 9, 9, 3, 3, 3, 3, 1, same, 1]
+  - [9, 16, 4, 4, 4, 128, 2, 2, 2, 1, same, 4]
+  - [32, 8, 8, 8, 8, 32, 3, 3, 3, 1, same, 1]
+conv_depthwise2d:
+  shape_desc: N, C, H, W, K_h, K_w, stride, padding, dilation
+  shapes:
+  - [32, 32, 64, 64, 3, 3, 1, 1, 1]
+  - [16, 64, 128, 128, 3, 3, 2, 1, 1]
+  - [8, 128, 56, 56, 5, 5, 1, 2, 1]
+  - [4, 256, 28, 28, 3, 3, 1, 1, 2]
+  - [2, 512, 14, 14, 3, 3, 1, 1, 1]
+  - [1, 64, 224, 224, 3, 3, 1, 1, 1]
+  - [1, 128, 112, 112, 5, 5, 2, 2, 1]
+conv_transpose1d:
+  shape_desc: batch, input_c, input_l, out_c, kernel, stride, padding, groups
+  shapes:
+  - [32, 64, 128, 64, 3, 1, 0, 1]
+  - [64, 48, 256, 128, 5, 2, 2, 1]
+  - [16, 24, 512, 96, 7, 1, 3, 1]
+  - [8, 16, 1024, 32, 3, 2, 1, 2]
+  - [4, 8, 2048, 16, 5, 1, 2, 1]
+conv_transpose2d:
+  shape_desc: N, C_in, H, W, C_out, K_h, K_w, stride, padding, groups
+  shapes:
+  - [1, 64, 128, 128, 64, 3, 3, 1, 1, 1]
+  - [1, 64, 64, 64, 32, 3, 3, 2, 1, 1]
+  - [4, 32, 32, 32, 32, 3, 3, 2, 1, 1]
+  - [8, 16, 64, 64, 16, 5, 5, 2, 2, 1]
+  - [16, 32, 16, 16, 64, 3, 3, 2, 1, 1]
+  - [16, 32, 32, 32, 64, 3, 3, 2, 1, 1]
+  - [32, 64, 32, 32, 32, 3, 3, 1, 0, 1]
+  - [32, 64, 16, 16, 32, 3, 3, 2, 1, 1]
+copy:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+copy_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+copysign:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+copysign_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+cos:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+cos_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+cosh:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+cosh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+cosh_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+count_nonzero:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+cp_gather_indexer_k_quant_cache:
+  shape_desc: batch_size, seq_len, block_size, head_dim, quant_block_size
+  shapes:
+  - [1, 1, 16, 512, 128]
+  - [1, 2, 16, 512, 128]
+  - [1, 4, 16, 512, 128]
+  - [1, 8, 16, 512, 128]
+  - [1, 16, 16, 512, 128]
+  - [1, 17, 16, 512, 128]
+  - [1, 32, 16, 512, 128]
+  - [1, 64, 16, 512, 128]
+  - [1, 128, 16, 512, 128]
+  - [1, 256, 16, 512, 128]
+  - [1, 512, 16, 512, 128]
+  - [1, 1024, 16, 512, 128]
+  - [1, 2048, 16, 512, 128]
+  - [1, 4096, 16, 512, 128]
+  - [1, 8192, 16, 512, 128]
+  - [1, 16384, 16, 512, 128]
+  - [1, 32768, 16, 512, 128]
+  - [1, 65536, 16, 512, 128]
+  - [2, 1, 16, 512, 128]
+  - [4, 1, 16, 512, 128]
+  - [8, 1, 16, 512, 128]
+  - [16, 1, 16, 512, 128]
+  - [32, 1, 16, 512, 128]
+  - [64, 1, 16, 512, 128]
+  - [128, 1, 16, 512, 128]
+  - [256, 1, 16, 512, 128]
+  - [2, 4096, 16, 512, 128]
+  - [4, 4096, 16, 512, 128]
+  - [8, 4096, 16, 512, 128]
+  - [16, 2048, 16, 512, 128]
+  - [32, 1024, 16, 512, 128]
+  - [64, 512, 16, 512, 128]
+  - [128, 256, 16, 512, 128]
+  - [256, 128, 16, 512, 128]
+  - [1, 8192, 64, 512, 128]
+  - [1, 32768, 64, 512, 128]
+  - [1, 65536, 64, 512, 128]
+  - [16, 2048, 64, 512, 128]
+  - [64, 512, 64, 512, 128]
+  - [256, 128, 64, 512, 128]
+cross_entropy_loss:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+ctc_loss:
+  shape_desc: T, N, C, S
+  shapes:
+  - [64, 4, 32, 16]
+  - [256, 16, 64, 48]
+  - [512, 32, 64, 48]
+  - [1024, 32, 128, 96]
+ctc_loss_backward:
+  shape_desc: T, N, C, S
+  shapes:
+  - [64, 4, 32, 16]
+  - [256, 16, 64, 48]
+  - [512, 32, 64, 48]
+  - [1024, 32, 128, 96]
+cudnn_convolution:
+  shape_desc: batch, input_c, input_h, input_w, out_c, kernel_h, kernel_w, stride,
+    padding, groups
+  shapes:
+  - [32, 64, 128, 128, 32, 3, 3, 1, 2, 1]
+  - [32, 64, 210, 210, 16, 5, 5, 2, 1, 1]
+  - [16, 32, 12, 12, 24, 3, 3, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 2, 2, 2]
+  - [16, 32, 24, 24, 24, 3, 3, 1, 2, 2]
+cummax:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+cummin:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+cumprod:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+cumprod_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+cumsum:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+cutlass_scaled_mm:
+  shape_desc: M, N, K
+  shapes:
+  - [1, 256, 128]
+  - [1, 16384, 1024]
+  - [1, 24576, 496]
+  - [16, 256, 496]
+  - [16, 16384, 128]
+  - [16, 24576, 4096]
+  - [32, 8192, 4096]
+  - [32, 16384, 4096]
+  - [33, 1024, 1024]
+  - [33, 8192, 128]
+  - [64, 2048, 496]
+  - [64, 16384, 1024]
+  - [100, 8192, 496]
+  - [128, 32768, 4096]
+  - [256, 4096, 4096]
+  - [512, 256, 1024]
+  - [512, 8192, 4096]
+  - [512, 16384, 128]
+  - [512, 24576, 128]
+dequantize_and_gather_k_cache:
+  shape_desc: M, N
+  shapes:
+  - [1, 512, 128, 512, 448, 64]
+  - [2, 1024, 256, 512, 448, 64]
+  - [4, 2048, 512, 512, 448, 64]
+  - [4, 2048, 2048, 512, 448, 64]
+  - [8, 4096, 1024, 512, 448, 64]
+dgeglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+diag:
+  shape_desc: M, N
+  shapes:
+  - [64]
+  - [1024]
+  - [64, 64]
+  - [256, 1024]
+  - [1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+diag_embed:
+  shape_desc: M, N
+  shapes:
+  - [4, 4]
+  - [16, 16]
+  - [128, 128]
+  - [256, 256]
+  - [1024, 1024]
+diagonal_backward:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+diff:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+digamma_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+div_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+div_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+dot:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+dreglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+dropout:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+dropout_backward:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+dswiglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+dunder_ior_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+dunder_ior_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+dunder_or_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+dunder_or_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+einsum:
+  shape_desc: B, M, N, K
+  shapes:
+  - [1, 512, 512, 512]
+  - [1, 1024, 1024, 1024]
+  - [16, 512, 512, 512]
+elu:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+elu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+elu_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+embedding:
+  shape_desc: M, N
+  shapes:
+  - [4, 4]
+  - [16, 16]
+  - [128, 128]
+  - [256, 256]
+  - [1024, 1024]
+embedding_backward:
+  shape_desc: M, N
+  shapes:
+  - [4, 4]
+  - [16, 16]
+  - [128, 128]
+  - [256, 256]
+  - [1024, 1024]
+embedding_dense_backward:
+  shape_desc: M, N
+  shapes:
+  - [32, 2048, 128, 8192]
+  - [16, 2048, 256, 16384]
+  - [8, 4096, 256, 32768]
+eq:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+eq_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+erf:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+erf_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+exp:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+exp2:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+exp2_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+exp_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+exp_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+expm1:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+expm1_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+expm1_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+exponential_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+eye:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+eye_m:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+feature_dropout:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+feature_dropout_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fft:
+  shape_desc: M, N
+  shapes:
+  - [128, 64]
+  - [128, 128]
+  - [128, 256]
+  - [128, 512]
+  - [128, 1024]
+  - [256, 64]
+  - [256, 128]
+  - [256, 256]
+  - [256, 512]
+  - [256, 1024]
+  - [512, 64]
+  - [512, 128]
+  - [512, 256]
+  - [512, 512]
+  - [512, 1024]
+  - [1024, 64]
+  - [1024, 128]
+  - [1024, 256]
+  - [1024, 512]
+  - [1024, 1024]
+  - [4096, 64]
+  - [4096, 128]
+  - [4096, 256]
+  - [4096, 512]
+  - [4096, 1024]
+fill_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fill_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fill_scalar_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fill_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fill_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fill_tensor_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+flash_mla:
+  shape_desc: M, N
+  shapes:
+  - [1024]
+  - [2048]
+  - [4096]
+  - [8192]
+  - [16384]
+flash_mla_sparse_fwd:
+  shape_desc: s_q, s_kv, topk, h_q, d_qk
+  shapes:
+  - [4096, 8192, 2048, 128, 576]
+  - [4096, 32768, 2048, 128, 576]
+  - [4096, 65536, 2048, 128, 576]
+  - [4096, 98304, 2048, 128, 576]
+  - [4096, 131072, 2048, 128, 576]
+  - [4096, 8192, 512, 64, 512]
+  - [4096, 32768, 512, 64, 512]
+  - [4096, 49152, 512, 64, 512]
+  - [4096, 65536, 512, 64, 512]
+  - [4096, 8192, 1024, 128, 512]
+  - [4096, 32768, 1024, 128, 512]
+  - [4096, 49152, 1024, 128, 512]
+  - [4096, 65536, 1024, 128, 512]
+flip:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+float_power_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+floor:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+floor_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+floor_divide_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+floor_divide_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+floor_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+fmin:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+fmin_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fmod_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+fmod_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fmod_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fmod_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fmod_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fp8_einsum:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 1024]
+  - [4, 8, 4096, 1024]
+  - [8, 8, 4096, 1024]
+  - [16, 8, 4096, 1024]
+  - [32, 8, 4096, 1024]
+  - [64, 8, 4096, 1024]
+  - [128, 8, 4096, 1024]
+  - [4096, 8, 4096, 1024]
+  - [8192, 8, 4096, 1024]
+  - [16384, 8, 4096, 1024]
+  - [32768, 8, 4096, 1024]
+  - [1, 16, 7168, 1024]
+  - [4, 16, 7168, 1024]
+  - [8, 16, 7168, 1024]
+  - [16, 16, 7168, 1024]
+  - [32, 16, 7168, 1024]
+  - [64, 16, 7168, 1024]
+  - [128, 16, 7168, 1024]
+  - [4096, 16, 7168, 1024]
+  - [8192, 16, 7168, 1024]
+  - [16384, 16, 7168, 1024]
+  - [32768, 16, 7168, 1024]
+fp8_paged_mqa_logits_gems_vs_deepgemm:
+  shape_desc: batch_size, next_n, heads, head_dim, avg_kv
+  shapes:
+  - [1, 1, 16, 64, 1024]
+  - [2, 1, 32, 128, 2048]
+  - [4, 1, 32, 128, 2048]
+  - [2, 2, 32, 128, 2048]
+  - [8, 1, 32, 128, 3072]
+full:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+full_like:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+functional_sym_constrain_range_for_size:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fused_add_rms_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert:
+  shape_desc: num_tokens, num_heads, num_tokens_insert, block_size, max_pos, eps
+  shapes:
+  - [1, 64, 1, 64, 4096, 1.0e-06]
+  - [1, 128, 1, 64, 4096, 1.0e-06]
+  - [4, 64, 4, 64, 4096, 1.0e-06]
+  - [4, 128, 4, 64, 4096, 1.0e-06]
+  - [17, 64, 17, 64, 4096, 1.0e-06]
+  - [17, 128, 17, 64, 4096, 1.0e-06]
+  - [64, 64, 64, 64, 4096, 1.0e-06]
+  - [64, 128, 64, 64, 4096, 1.0e-06]
+  - [1024, 64, 1024, 64, 4096, 1.0e-06]
+  - [1024, 128, 1024, 64, 4096, 1.0e-06]
+  - [2048, 64, 2048, 64, 4096, 1.0e-06]
+  - [2048, 128, 2048, 64, 4096, 1.0e-06]
+  - [8192, 64, 8192, 64, 4096, 1.0e-06]
+  - [8192, 128, 8192, 64, 4096, 1.0e-06]
+  - [32768, 64, 32768, 64, 4096, 1.0e-06]
+  - [32768, 128, 32768, 64, 4096, 1.0e-06]
+  - [65536, 64, 65536, 64, 4096, 1.0e-06]
+  - [65536, 128, 65536, 64, 4096, 1.0e-06]
+  - [98304, 64, 98304, 64, 4096, 1.0e-06]
+  - [98304, 128, 98304, 64, 4096, 1.0e-06]
+  - [131072, 64, 131072, 64, 4096, 1.0e-06]
+  - [131072, 128, 131072, 64, 4096, 1.0e-06]
+fused_experts_impl:
+  shape_desc: num_tokens, num_experts, hidden_size, intermediate_size, topk
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+  - [4, 8, 4096, 14336, 2]
+  - [16, 8, 4096, 14336, 2]
+  - [64, 8, 4096, 14336, 2]
+  - [128, 8, 4096, 14336, 2]
+  - [256, 8, 4096, 14336, 2]
+  - [512, 8, 4096, 14336, 2]
+  - [1, 256, 7168, 2048, 8]
+  - [4, 256, 7168, 2048, 8]
+  - [16, 256, 7168, 2048, 8]
+  - [64, 256, 7168, 2048, 8]
+  - [128, 256, 7168, 2048, 8]
+  - [256, 256, 7168, 2048, 8]
+  - [1, 512, 4096, 1024, 10]
+  - [4, 512, 4096, 1024, 10]
+  - [16, 512, 4096, 1024, 10]
+  - [64, 512, 4096, 1024, 10]
+  - [128, 512, 4096, 1024, 10]
+  - [256, 512, 4096, 1024, 10]
+fused_inv_rope_fp8_quant:
+  shape_desc: num_tokens, num_heads, n_groups, tma_aligned_scales
+  shapes:
+  - [256, 128, 8, false]
+  - [512, 128, 8, false]
+  - [1024, 128, 8, false]
+  - [2048, 128, 8, false]
+  - [4096, 128, 8, false]
+  - [256, 128, 8, true]
+  - [512, 128, 8, true]
+  - [1024, 128, 8, true]
+  - [2048, 128, 8, true]
+  - [4096, 128, 8, true]
+fused_marlin_moe:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+  - [4, 8, 4096, 14336, 2]
+  - [8, 8, 4096, 14336, 2]
+  - [16, 8, 4096, 14336, 2]
+  - [32, 8, 4096, 14336, 2]
+  - [64, 8, 4096, 14336, 2]
+  - [128, 8, 4096, 14336, 2]
+  - [256, 8, 4096, 14336, 2]
+  - [1, 256, 7168, 2048, 8]
+  - [4, 256, 7168, 2048, 8]
+  - [8, 256, 7168, 2048, 8]
+  - [16, 256, 7168, 2048, 8]
+  - [32, 256, 7168, 2048, 8]
+  - [64, 256, 7168, 2048, 8]
+  - [128, 256, 7168, 2048, 8]
+  - [256, 256, 7168, 2048, 8]
+  - [1, 512, 4096, 1024, 10]
+  - [4, 512, 4096, 1024, 10]
+  - [8, 512, 4096, 1024, 10]
+  - [16, 512, 4096, 1024, 10]
+  - [32, 512, 4096, 1024, 10]
+  - [64, 512, 4096, 1024, 10]
+  - [128, 512, 4096, 1024, 10]
+  - [256, 512, 4096, 1024, 10]
+  - [1, 256, 4096, 2048, 6]
+  - [4, 256, 4096, 2048, 6]
+  - [8, 256, 4096, 2048, 6]
+  - [16, 256, 4096, 2048, 6]
+  - [32, 256, 4096, 2048, 6]
+  - [64, 256, 4096, 2048, 6]
+  - [128, 256, 4096, 2048, 6]
+  - [256, 256, 4096, 2048, 6]
+fused_marlin_moe_int8:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+  - [16, 8, 4096, 14336, 2]
+  - [64, 8, 4096, 14336, 2]
+  - [1, 256, 7168, 2048, 8]
+  - [16, 256, 7168, 2048, 8]
+  - [64, 256, 7168, 2048, 8]
+fused_moe:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+  - [4, 8, 4096, 14336, 2]
+  - [16, 8, 4096, 14336, 2]
+  - [64, 8, 4096, 14336, 2]
+  - [128, 8, 4096, 14336, 2]
+  - [256, 8, 4096, 14336, 2]
+  - [512, 8, 4096, 14336, 2]
+  - [1, 64, 7168, 2048, 8]
+  - [4, 64, 7168, 2048, 8]
+  - [16, 64, 7168, 2048, 8]
+  - [64, 64, 7168, 2048, 8]
+  - [128, 64, 7168, 2048, 8]
+  - [256, 64, 7168, 2048, 8]
+  - [1, 128, 4096, 1024, 10]
+  - [4, 128, 4096, 1024, 10]
+  - [16, 128, 4096, 1024, 10]
+  - [64, 128, 4096, 1024, 10]
+  - [128, 128, 4096, 1024, 10]
+  - [256, 128, 4096, 1024, 10]
+fused_moe_int8_w8a16_gems_vs_bf16_deq:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+  - [4, 8, 4096, 14336, 2]
+  - [16, 8, 4096, 14336, 2]
+  - [64, 8, 4096, 14336, 2]
+  - [128, 8, 4096, 14336, 2]
+  - [256, 8, 4096, 14336, 2]
+  - [512, 8, 4096, 14336, 2]
+  - [1, 256, 7168, 2048, 8]
+  - [4, 256, 7168, 2048, 8]
+  - [16, 256, 7168, 2048, 8]
+  - [64, 256, 7168, 2048, 8]
+  - [128, 256, 7168, 2048, 8]
+  - [256, 256, 7168, 2048, 8]
+fused_q_kv_rmsnorm:
+  shape_desc: M, N
+  shapes:
+  - [1, 1536, 512]
+  - [32, 1536, 512]
+  - [128, 1536, 512]
+  - [512, 1536, 512]
+  - [2048, 1536, 512]
+  - [32, 36864, 576]
+  - [128, 36864, 576]
+gather:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+gather_backward:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+gcd:
+  shape_desc: (B), M, N
+  shapes:
+  - [65536]
+  - [1024, 1024]
+  - [256, 4096]
+  - [64, 256, 256]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+gcd_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+ge:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+ge_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+geglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+gelu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+gelu_and_mul:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+gelu_backward:
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [256, 1024, 1024]
+glu:
+  shape_desc: M, N
+  shapes:
+  - [4, 8, 512, 128]
+  - [4, 8, 1024, 128]
+  - [4, 8, 2048, 128]
+  - [4, 8, 3072, 128]
+  - [4, 8, 4096, 128]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+glu_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+greater:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+greater_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+greater_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+greater_scalar_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+grid_sample:
+  shape_desc: M, N
+  shapes:
+  - [1, 3, 32, 32]
+  - [2, 16, 32, 32]
+  - [1, 3, 8, 8, 8]
+  - [2, 4, 8, 8, 8]
+  - [2, 32, 64, 64]
+  - [4, 64, 64, 64]
+  - [2, 8, 16, 16, 16]
+  - [2, 16, 16, 16, 16]
+  - [4, 128, 128, 128]
+  - [2, 32, 32, 32, 32]
+group_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+  - [16, 16, 64]
+  - [16, 16, 1024]
+  - [16, 16, 4098]
+  - [1, 8, 4, 4]
+  - [16, 8, 128, 128]
+grouped_mm:
+  shape_desc: Groups, N, K
+  shapes:
+  - [16, 512, 2048]
+  - [16, 2560, 2048]
+  - [64, 2048, 128]
+grouped_topk:
+  shape_desc: M, N
+  shapes:
+  - [1, 64, 8, 2, 8]
+  - [8, 64, 8, 2, 8]
+  - [32, 64, 8, 2, 8]
+  - [64, 64, 8, 2, 8]
+  - [128, 64, 8, 2, 8]
+  - [256, 64, 8, 2, 8]
+  - [32, 128, 8, 2, 8]
+  - [64, 128, 8, 2, 8]
+  - [128, 128, 8, 2, 8]
+  - [64, 64, 4, 2, 4]
+  - [64, 128, 16, 2, 8]
+  - [512, 64, 8, 2, 8]
+  - [1024, 64, 8, 2, 8]
+  - [2048, 64, 8, 2, 8]
+groupmm:
+  shape_desc: Groups, N, K
+  shapes:
+  - [16, 512, 2048]
+  - [16, 2560, 2048]
+  - [64, 2048, 128]
+gt:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+gt_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+hadamard_transform:
+  shape_desc: batch, dim
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+hadamard_transform_mn:
+  shape_desc: batch, dim
+  shapes:
+  - [1024, 1536]
+  - [1024, 3072]
+  - [1024, 6144]
+  - [1024, 12288]
+  - [8192, 1536]
+  - [8192, 3072]
+  - [8192, 6144]
+  - [8192, 12288]
+  - [32768, 1536]
+  - [32768, 3072]
+  - [32768, 6144]
+  - [32768, 12288]
+  - [1024, 10240]
+  - [1024, 14336]
+  - [1024, 20480]
+  - [8192, 10240]
+  - [8192, 14336]
+  - [8192, 20480]
+  - [1024, 40960]
+  - [8192, 40960]
+hardsigmoid:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+hardswish_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+hc_head_fused_kernel:
+  shape_desc: N, hidden_size
+  shapes:
+  - [1, 1280, 4]
+  - [4, 2560, 4]
+  - [16, 4096, 4]
+  - [64, 7168, 4]
+  - [256, 1280, 2]
+  - [256, 1280, 4]
+  - [512, 1280, 2]
+  - [512, 1280, 4]
+  - [512, 2560, 2]
+  - [512, 2560, 4]
+  - [1024, 2560, 2]
+  - [1024, 2560, 4]
+  - [2048, 4096, 2]
+  - [2048, 4096, 4]
+  - [4096, 4096, 2]
+  - [4096, 4096, 4]
+hc_split_sinkhorn_forward:
+  shape_desc: batch, seqlen
+  shapes:
+  - [8, 16, 4]
+  - [32, 64, 4]
+  - [128, 128, 4]
+  - [256, 256, 4]
+  - [8, 16, 2]
+  - [32, 64, 2]
+  - [128, 128, 2]
+  - [256, 256, 2]
+histc:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+hstack:
+  shapes:
+  - [536870912]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [512, 512, 512]
+huber_loss:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [2048, 32768]
+  - [2048, 65536]
+  - [4096, 32768]
+  - [4096, 65536]
+  - [3072, 49152]
+  - [3072, 65536]
+hypot:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+hypot_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+i0:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+i0_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+i0_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+index:
+  shape_desc: M, N
+  shapes:
+  - - [268435456]
+    - - [65536]
+  - - [32, 32]
+    - - [8]
+      - [8]
+  - - [32, 32]
+    - - [8]
+      - [2, 8]
+  - - [32, 32]
+    - - [2, 8]
+  - - [1024, 1024]
+    - - [64]
+      - [64]
+  - - [512, 512, 512]
+    - - [128]
+      - [128]
+      - [128]
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [2, 128]
+      - [2, 128]
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [128]
+      - [128]
+  - - [512, 512, 512]
+    - - [2, 128]
+  - - [64, 64, 64]
+    - - [2, 8]
+      - [2, 8]
+index_add:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+index_add_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+index_copy:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1, 2]
+  - [4096, 256]
+  - [200, 40999, 3]
+index_copy_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1, 2]
+  - [4096, 256]
+  - [200, 40999, 3]
+index_put:
+  shape_desc: input_shape, [index_tensor_shapes], values_shape, accumulate
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [8]
+    - [8]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [2, 8]
+    - [8]
+    - false
+  - - [32, 32]
+    - - [2, 8]
+    - [32]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [64]
+    - [64]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [4, 64]
+    - [64]
+    - false
+  - - [1024, 1024]
+    - - [4, 64]
+    - [1024]
+    - false
+  - - [512, 512, 512]
+    - - [128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+    - [512]
+    - false
+  - - [100]
+    - - [100]
+    - [100]
+    - true
+  - - [32, 32]
+    - - [32, 32]
+    - [32, 32]
+    - true
+  - - [16, 16, 4]
+    - - [16, 16, 4]
+    - [16, 16, 4]
+    - true
+  - - [1024, 1024]
+    - - [1024, 1024]
+    - [1048576]
+    - true
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [2, 128]
+      - [2, 128]
+    - [2, 128]
+    - false
+  - - [512, 512]
+    - - [512, 512]
+    - [262144]
+    - true
+  - - [64, 64, 64]
+    - - [64, 64, 64]
+    - [262144]
+    - true
+index_put_:
+  shape_desc: input_shape, [index_tensor_shapes], values_shape, accumulate
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [8]
+    - [8]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [2, 8]
+    - [8]
+    - false
+  - - [32, 32]
+    - - [2, 8]
+    - [32]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [64]
+    - [64]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [4, 64]
+    - [64]
+    - false
+  - - [1024, 1024]
+    - - [4, 64]
+    - [1024]
+    - false
+  - - [512, 512, 512]
+    - - [128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+    - [512]
+    - false
+  - - [100]
+    - - [100]
+    - [100]
+    - true
+  - - [32, 32]
+    - - [32, 32]
+    - [32, 32]
+    - true
+  - - [16, 16, 4]
+    - - [16, 16, 4]
+    - [16, 16, 4]
+    - true
+  - - [1024, 1024]
+    - - [1024, 1024]
+    - [1048576]
+    - true
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [2, 128]
+      - [2, 128]
+    - [2, 128]
+    - false
+  - - [512, 512]
+    - - [512, 512]
+    - [262144]
+    - true
+  - - [64, 64, 64]
+    - - [64, 64, 64]
+    - [262144]
+    - true
+index_put_impl:
+  shape_desc: input_shape, [index_tensor_shapes], values_shape, accumulate
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [8]
+    - [8]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [2, 8]
+    - [8]
+    - false
+  - - [32, 32]
+    - - [2, 8]
+    - [32]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [64]
+    - [64]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [4, 64]
+    - [64]
+    - false
+  - - [1024, 1024]
+    - - [4, 64]
+    - [1024]
+    - false
+  - - [512, 512, 512]
+    - - [128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+    - [512]
+    - false
+  - - [100]
+    - - [100]
+    - [100]
+    - true
+  - - [32, 32]
+    - - [32, 32]
+    - [32, 32]
+    - true
+  - - [16, 16, 4]
+    - - [16, 16, 4]
+    - [16, 16, 4]
+    - true
+  - - [1024, 1024]
+    - - [1024, 1024]
+    - [1048576]
+    - true
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [2, 128]
+      - [2, 128]
+    - [2, 128]
+    - false
+  - - [512, 512]
+    - - [512, 512]
+    - [262144]
+    - true
+  - - [64, 64, 64]
+    - - [64, 64, 64]
+    - [262144]
+    - true
+index_select:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+index_select_backward:
+  shape_desc: N, M
+  shapes:
+  - [16, 64]
+  - [32, 128]
+  - [64, 256]
+  - [16, 16, 64]
+  - [32, 32, 128]
+indexer_k_quant_and_cache:
+  shape_desc: num_tokens, num_blocks, block_size, head_dim, quant_block_size
+  shapes:
+  - [1, 1, 16, 512, 128]
+  - [2, 1, 16, 512, 128]
+  - [4, 1, 16, 512, 128]
+  - [8, 1, 16, 512, 128]
+  - [16, 2, 16, 512, 128]
+  - [17, 3, 16, 512, 128]
+  - [32, 4, 16, 512, 128]
+  - [64, 8, 16, 512, 128]
+  - [128, 16, 16, 512, 128]
+  - [256, 32, 16, 512, 128]
+  - [512, 64, 16, 512, 128]
+  - [1024, 128, 16, 512, 128]
+  - [2048, 256, 16, 512, 128]
+  - [4096, 512, 16, 512, 128]
+  - [8192, 1024, 16, 512, 128]
+  - [16384, 2048, 16, 512, 128]
+  - [32768, 4096, 16, 512, 128]
+  - [65536, 8192, 16, 512, 128]
+  - [8192, 256, 64, 512, 128]
+  - [32768, 1024, 64, 512, 128]
+  - [65536, 2048, 64, 512, 128]
+inplace_fused_experts:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+  - [4, 8, 4096, 14336, 2]
+  - [16, 8, 4096, 14336, 2]
+  - [64, 8, 4096, 14336, 2]
+  - [128, 8, 4096, 14336, 2]
+  - [256, 8, 4096, 14336, 2]
+  - [512, 8, 4096, 14336, 2]
+  - [1, 256, 7168, 2048, 8]
+  - [4, 256, 7168, 2048, 8]
+  - [16, 256, 7168, 2048, 8]
+  - [64, 256, 7168, 2048, 8]
+  - [128, 256, 7168, 2048, 8]
+  - [256, 256, 7168, 2048, 8]
+instance_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+  - [16, 16, 64]
+  - [16, 16, 1024]
+  - [16, 16, 4098]
+  - [1, 8, 4, 4]
+  - [16, 8, 128, 128]
+is_all_true:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+isclose:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+isfinite:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+isin:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+isin_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+isin_tensor_scalar:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+isinf:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+isnan:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+isneginf:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+kron:
+  shape_desc: M, N
+  shapes:
+  - [16, 16]
+  - [64, 64]
+  - [128, 128]
+  - [256, 256]
+  - [4, 8, 16, 32]
+layer_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+  - [16, 16, 64]
+  - [16, 16, 1024]
+  - [16, 16, 4098]
+  - [1, 8, 4, 4]
+  - [16, 8, 128, 128]
+layer_norm_backward:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+  - [16, 16, 64]
+  - [16, 16, 1024]
+  - [16, 16, 4098]
+  - [1, 8, 4, 4]
+  - [16, 8, 128, 128]
+le:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+le_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+leaky_relu:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+leaky_relu_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+lerp_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+lerp_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+lerp_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+lerp_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+lift_fresh_copy:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+linspace:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+log:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log10:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log10_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log10_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log1p:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log1p_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log_sigmoid:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+log_softmax:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+log_softmax_backward_data:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+log_softmax_backward_data_out:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+log_softmax_out:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+logaddexp:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+logaddexp_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+logical_and:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+logical_and_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+logical_not:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+logical_or:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+logical_or_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+logical_xor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+logit:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+logit_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+logspace:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+logsumexp:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+lt:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+lt_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+margin_ranking_loss:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1048576]
+  - [1024, 1]
+  - [1024, 256]
+  - [1024, 4096]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+masked_fill:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+masked_fill_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+masked_fill_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+masked_fill_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+masked_scatter:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+masked_scatter_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+masked_select:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+max:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+max_dim:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+max_pool2d_backward:
+  shape_desc: N, C, H, W
+  shapes:
+  - [4, 3, 224, 224]
+  - [16, 64, 56, 56]
+  - [32, 128, 28, 28]
+  - [64, 256, 14, 14]
+  - [128, 512, 7, 7]
+max_pool2d_with_indices:
+  shape_desc: N, C, H, W
+  shapes:
+  - [4, 3, 224, 224]
+  - [16, 64, 56, 56]
+  - [32, 128, 28, 28]
+  - [64, 256, 14, 14]
+  - [128, 512, 7, 7]
+max_pool3d:
+  shape_desc: N, C, D, H, W
+  shapes:
+  - [4, 3, 16, 56, 56]
+  - [8, 64, 8, 28, 28]
+  - [16, 128, 4, 14, 14]
+  - [32, 256, 2, 7, 7]
+max_pool3d_backward:
+  shape_desc: N, C, D, H, W
+  shapes:
+  - [4, 3, 16, 56, 56]
+  - [8, 64, 8, 28, 28]
+  - [16, 128, 4, 14, 14]
+  - [32, 256, 2, 7, 7]
+maximum:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+mean:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+mean_dim:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+median:
+  shape_desc: input shape
+  shapes:
+  - [2]
+  - [64]
+  - [65]
+  - [128]
+  - [256]
+  - [257]
+  - [1024]
+  - [4096]
+  - [4, 8, 16, 8]
+median_dim:
+  shape_desc: input shape or [input shape, dim, keepdim]
+  shapes:
+  - [1536, 640]
+  - [144, 8192]
+  - [768, 1536]
+  - [32, 384, 128]
+  - - [32, 384, 128]
+    - 1
+    - false
+  - - [640, 1536]
+    - 0
+    - false
+  - - [8, 64, 384, 64]
+    - 2
+    - false
+mhc_bwd:
+  shape_desc: seqlen, n_stream
+  shapes:
+  - [256, 4]
+  - [1024, 4]
+  - [4096, 4]
+  - [8192, 4]
+  - [16384, 4]
+  - [65536, 4]
+mhc_post:
+  shape_desc: N, H
+  shapes:
+  - [4096, 1280]
+  - [4096, 2560]
+  - [4096, 7168]
+mhc_pre:
+  shape_desc: N, hidden_size
+  shapes:
+  - [512, 1280]
+  - [512, 2560]
+  - [512, 4096]
+  - [1024, 1280]
+  - [1024, 2560]
+  - [1024, 4096]
+  - [2048, 1280]
+  - [2048, 2560]
+  - [2048, 4096]
+  - [8192, 1280]
+  - [8192, 2560]
+  - [8192, 4096]
+min:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+min_dim:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+minimum:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+mm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+mm_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+  - [8, 1848, 1536, 151936]
+  - [8, 1848, 1536, 128256]
+  - [8, 1848, 1536, 152064]
+  - [8, 4096, 1, 152064]
+  - [4, 1, 1024, 4096]
+  - [4, 1, 128256, 4096]
+  - [4, 1, 14336, 4096]
+  - [4, 1, 4096, 14336]
+  - [4, 1, 4096, 4096]
+  - [4, 1, 6144, 4096]
+  - [4, 1, 28672, 4096]
+  - [4, 1, 3584, 3584]
+  - [4, 1, 18944, 3584]
+  - [4, 1, 3584, 18944]
+  - [4, 1, 152064, 3584]
+  - [4, 1, 37888, 3584]
+  - [4, 1, 512, 3584]
+  - [4, 1, 4608, 3584]
+  - [4, 2, 1024, 4096]
+  - [4, 2, 128256, 4096]
+  - [4, 2, 14336, 4096]
+  - [4, 2, 4096, 14336]
+  - [4, 2, 4096, 4096]
+  - [4, 2, 6144, 4096]
+  - [4, 2, 28672, 4096]
+  - [4, 2, 3584, 3584]
+  - [4, 2, 18944, 3584]
+  - [4, 2, 3584, 18944]
+  - [4, 2, 152064, 3584]
+  - [4, 2, 37888, 3584]
+  - [4, 2, 512, 3584]
+  - [4, 2, 4608, 3584]
+  - [4, 3, 1024, 4096]
+  - [4, 3, 128256, 4096]
+  - [4, 3, 14336, 4096]
+  - [4, 3, 4096, 14336]
+  - [4, 3, 4096, 4096]
+  - [4, 3, 6144, 4096]
+  - [4, 3, 28672, 4096]
+  - [4, 3, 3584, 3584]
+  - [4, 3, 18944, 3584]
+  - [4, 3, 3584, 18944]
+  - [4, 3, 152064, 3584]
+  - [4, 3, 37888, 3584]
+  - [4, 3, 512, 3584]
+  - [4, 3, 4608, 3584]
+  - [4, 4, 1024, 4096]
+  - [4, 4, 128256, 4096]
+  - [4, 4, 14336, 4096]
+  - [4, 4, 4096, 14336]
+  - [4, 4, 4096, 4096]
+  - [4, 4, 6144, 4096]
+  - [4, 4, 28672, 4096]
+  - [4, 4, 3584, 3584]
+  - [4, 4, 18944, 3584]
+  - [4, 4, 3584, 18944]
+  - [4, 4, 152064, 3584]
+  - [4, 4, 37888, 3584]
+  - [4, 4, 512, 3584]
+  - [4, 4, 4608, 3584]
+  - [4, 8, 1024, 4096]
+  - [4, 8, 128256, 4096]
+  - [4, 8, 14336, 4096]
+  - [4, 8, 4096, 14336]
+  - [4, 8, 4096, 4096]
+  - [4, 8, 6144, 4096]
+  - [4, 8, 28672, 4096]
+  - [4, 8, 3584, 3584]
+  - [4, 8, 18944, 3584]
+  - [4, 8, 3584, 18944]
+  - [4, 8, 152064, 3584]
+  - [4, 8, 37888, 3584]
+  - [4, 8, 512, 3584]
+  - [4, 8, 4608, 3584]
+  - [4, 98, 1024, 4096]
+  - [4, 98, 128256, 4096]
+  - [4, 98, 14336, 4096]
+  - [4, 98, 4096, 14336]
+  - [4, 98, 4096, 4096]
+  - [4, 98, 6144, 4096]
+  - [4, 98, 28672, 4096]
+  - [4, 98, 3584, 3584]
+  - [4, 98, 18944, 3584]
+  - [4, 98, 3584, 18944]
+  - [4, 98, 152064, 3584]
+  - [4, 98, 37888, 3584]
+  - [4, 98, 512, 3584]
+  - [4, 98, 4608, 3584]
+  - [4, 256, 1024, 4096]
+  - [4, 256, 128256, 4096]
+  - [4, 256, 14336, 4096]
+  - [4, 256, 4096, 14336]
+  - [4, 256, 4096, 4096]
+  - [4, 256, 6144, 4096]
+  - [4, 256, 28672, 4096]
+  - [4, 256, 3584, 3584]
+  - [4, 256, 18944, 3584]
+  - [4, 256, 3584, 18944]
+  - [4, 256, 152064, 3584]
+  - [4, 256, 37888, 3584]
+  - [4, 256, 512, 3584]
+  - [4, 256, 4608, 3584]
+  - [4, 8192, 1024, 4096]
+  - [4, 8192, 128256, 4096]
+  - [4, 8192, 14336, 4096]
+  - [4, 8192, 4096, 14336]
+  - [4, 8192, 4096, 4096]
+  - [4, 8192, 6144, 4096]
+  - [4, 8192, 28672, 4096]
+  - [4, 8192, 3584, 3584]
+  - [4, 8192, 18944, 3584]
+  - [4, 8192, 3584, 18944]
+  - [4, 8192, 152064, 3584]
+  - [4, 8192, 37888, 3584]
+  - [4, 8192, 512, 3584]
+  - [4, 8192, 4608, 3584]
+mm_self_transpose:
+  shape_desc: M, K
+  shapes:
+  - [257, 96]
+  - [384, 384]
+  - [1536, 320]
+  - [3072, 768]
+  - [1024, 1024]
+  - [4096, 1024]
+  - [5333, 71]
+  - [8192, 2048]
+  - [10240, 4096]
+moe_align_block_size_triton:
+  shape_desc: M, N
+  shapes:
+  - [512, 64, 16384, 10]
+  - [512, 64, 6152, 10]
+  - [512, 64, 4727, 10]
+  - [512, 64, 1905, 10]
+  - [512, 64, 11575, 10]
+  - [512, 64, 1032, 10]
+  - [512, 64, 4201, 10]
+  - [512, 64, 2056, 10]
+  - [512, 64, 7561, 10]
+  - [512, 64, 4104, 10]
+  - [512, 64, 14281, 10]
+moe_sum:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+mse_loss:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+mul:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+mul_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+multinomial:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+mv:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+nan_to_num:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+nanmedian:
+  shape_desc: N or M, N
+  shapes:
+  - [64]
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+nanmedian_dim:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+nanmedian_dim_values:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+nanmedian_out:
+  shape_desc: N or M, N
+  shapes:
+  - [64]
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+ne:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+ne_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+neg:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+neg_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+new_full:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+nll_loss2d_backward:
+  shape_desc: M, N
+  shapes:
+  - [64, 64, 4, 8]
+  - [256, 256, 4, 8]
+  - [1024, 1024, 4, 8]
+  - [4096, 4096, 4, 8]
+  - [1024, 65536, 4, 8]
+nll_loss2d_forward:
+  shape_desc: M, N
+  shapes:
+  - [64, 64, 4, 8]
+  - [256, 256, 4, 8]
+  - [1024, 1024, 4, 8]
+  - [4096, 4096, 4, 8]
+  - [1024, 65536, 4, 8]
+nll_loss_backward:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+nll_loss_forward:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+nll_loss_nd_backward:
+  shape_desc: N, C, d1, d2, ...
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [10000, 65536]
+  - [32, 128, 512]
+  - [64, 64, 4, 8]
+  - [256, 256, 4, 8]
+  - [4096, 4096, 4, 8]
+  - [64, 64, 8, 4, 8]
+nll_loss_nd_forward:
+  shape_desc: N, C, d1, d2, ...
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [10000, 65536]
+  - [32, 128, 512]
+  - [64, 64, 4, 8]
+  - [256, 256, 4, 8]
+  - [4096, 4096, 4, 8]
+  - [64, 64, 8, 4, 8]
+nonzero:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+nonzero_numpy:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+normal_float_float_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+normal_float_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+normal_tensor_float:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+normal_tensor_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+one_hot:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+ones:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+ones_like:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+outer:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+pack_seq_triton:
+  shape_desc: N, D, B, lengths
+  shapes:
+  - - 512
+    - 64
+    - 5
+    - [64, 128, 64, 128, 128]
+  - - 4096
+    - 128
+    - 4
+    - [1024, 1024, 1024, 1024]
+  - - 8192
+    - 256
+    - 5
+    - [1024, 2048, 1024, 2048, 2048]
+  - - 2048
+    - 512
+    - 4
+    - [512, 512, 512, 512]
+  - - 16384
+    - 64
+    - 8
+    - [2048, 2048, 2048, 2048, 2048, 2048, 2048, 2048]
+  - - 1024
+    - 1024
+    - 4
+    - [256, 256, 256, 256]
+pad:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+per_token_group_quant_fp8:
+  shape_desc: M, N
+  shapes:
+  - [7, 512, 512]
+  - [7, 4096, 256]
+  - [83, 512, 64]
+  - [2048, 4096, 256]
+  - [2048, 13824, 512]
+pixel_shuffle:
+  shape_desc: N, C, H, W, upscale_factor
+  shapes:
+  - [1, 4, 2, 3, 2]
+  - [2, 9, 4, 4, 3]
+  - [4, 64, 32, 32, 2]
+  - [2, 128, 64, 64, 2]
+  - [1, 64, 16, 16, 4]
+  - [8, 36, 64, 64, 3]
+  - [1, 16, 128, 128, 2]
+pixel_unshuffle:
+  shape_desc: M, N
+  shapes:
+  - - [1, 3, 8, 8]
+    - 2
+  - - [2, 4, 12, 6]
+    - 3
+  - - [4, 16, 64, 48]
+    - 4
+poisson:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+polar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+pow_tensor_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+pow_tensor_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+prelu:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+prod:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+quantile:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 256]
+  - [64, 512, 128]
+  - [20, 8, 10000]
+  - [4]
+  - [1024]
+  - [65535]
+  - [1024, 1]
+  - [1024, 8]
+  - [1024, 64]
+  - [1024, 512]
+  - [1024, 4096]
+  - [64, 64, 1]
+  - [64, 64, 8]
+  - [64, 64, 64]
+  - [64, 64, 512]
+  - [64, 64, 4096]
+rad2deg:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+rad2deg_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+rand:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+rand_like:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+randint:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 512]
+  - [16, 131072]
+  - [8, 262144]
+randint_like:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+randn:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+randn_like:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+randperm:
+  shape_desc: M
+  shapes:
+  - [64]
+  - [1024]
+  - [2048]
+  - [4096]
+  - [65536]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+reciprocal:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+reciprocal_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+reflection_pad1d:
+  shape_desc: M, N
+  shapes:
+  - - [3, 33]
+    - [1, 1]
+  - - [2, 4, 64]
+    - [3, 5]
+  - - [8, 16, 256]
+    - [8, 8]
+  - - [32, 64, 2048]
+    - [3, 5]
+reflection_pad1d_backward:
+  shape_desc: M, N
+  shapes:
+  - [2, 3]
+  - [4, 8]
+  - [8, 16]
+  - [1, 32]
+  - [4, 64]
+  - [8, 128]
+reflection_pad1d_out:
+  shape_desc: M, N
+  shapes:
+  - - [3, 33]
+    - [1, 1]
+  - - [2, 4, 64]
+    - [3, 5]
+  - - [8, 16, 256]
+    - [8, 8]
+  - - [32, 64, 2048]
+    - [3, 5]
+reflection_pad2d:
+  shape_desc: M, N
+  shapes:
+  - - [3, 33, 33]
+    - [1, 1, 1, 1]
+  - - [2, 4, 32, 64]
+    - [2, 3, 2, 3]
+  - - [8, 16, 64, 64]
+    - [3, 5, 3, 5]
+  - - [32, 64, 128, 256]
+    - [0, 4, 0, 4]
+  - - [16, 32, 64, 128]
+    - [1, 1, 1, 1]
+reflection_pad2d_out:
+  shape_desc: M, N
+  shapes:
+  - - [3, 33, 33]
+    - [1, 1, 1, 1]
+  - - [2, 4, 32, 64]
+    - [2, 3, 2, 3]
+  - - [8, 16, 64, 64]
+    - [3, 5, 3, 5]
+  - - [32, 64, 128, 256]
+    - [0, 4, 0, 4]
+  - - [16, 32, 64, 128]
+    - [1, 1, 1, 1]
+reglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+relu:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+relu6:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+relu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+remainder:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+remainder_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+remainder_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+remainder_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+remainder_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+renorm:
+  shape_desc: M, N
+  shapes:
+  - [4, 8]
+  - [8, 16]
+  - [16, 32]
+  - [32, 64]
+  - [64, 128]
+  - [128, 256]
+  - [16, 32, 64]
+  - [8, 16, 128]
+repeat:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [16, 256, 256]
+  - [512, 512, 512]
+  - [64, 64, 64, 64]
+repeat_interleave_self_int:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [16, 256, 256]
+  - [512, 512, 512]
+  - [64, 64, 64, 64]
+repeat_interleave_self_tensor:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 256, 256]
+  - [128, 256, 256]
+  - [64, 64, 64, 64]
+repeat_interleave_tensor:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 256, 256]
+  - [128, 256, 256]
+  - [64, 64, 64, 64]
+replication_pad1d:
+  shape_desc: M, N
+  shapes:
+  - - [2, 3, 7]
+    - [1, 2]
+  - - [4, 16, 64]
+    - [3, 1]
+  - - [8, 32, 256]
+    - [1, 2]
+  - - [32, 256]
+    - [3, 1]
+replication_pad3d:
+  shape_desc: M, N
+  shapes:
+  - [1, 3, 16, 256, 256]
+  - [4, 16, 32, 64, 64]
+  - [8, 64, 8, 32, 32]
+  - [2, 32, 16, 128, 128]
+  - [1, 1, 64, 128, 128]
+reshape_and_cache:
+  shape_desc: M, N
+  shapes:
+  - [21, 4, 64, 8, 512]
+  - [42, 8, 64, 8, 1024]
+  - [42, 8, 80, 16, 10000]
+  - [42, 8, 120, 32, 10000]
+  - [42, 8, 256, 8, 1024]
+reshape_and_cache_flash:
+  shape_desc: M, N
+  shapes:
+  - [21, 4, 64, 8, 512]
+  - [21, 4, 64, 8, 1024]
+  - [21, 4, 80, 16, 10000]
+  - [42, 8, 64, 8, 1024]
+  - [42, 8, 80, 16, 10000]
+resolve_conj:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+resolve_neg:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+rms_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+roll:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+rot90:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [128, 128]
+  - [256, 256]
+  - [512, 512]
+  - [1024, 1024]
+  - [2048, 2048]
+  - [100, 200]
+  - [200, 400]
+  - [400, 800]
+round:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+round_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+round_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+router_gemm:
+  shape_desc: M, N, K
+  shapes:
+  - [1, 8, 4096]
+  - [16, 8, 4096]
+  - [64, 8, 4096]
+  - [256, 8, 4096]
+  - [1024, 8, 4096]
+  - [4096, 8, 4096]
+  - [64, 64, 7168]
+  - [256, 64, 7168]
+  - [1024, 128, 7168]
+  - [4096, 128, 7168]
+rrelu_with_noise_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+rsqrt:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+rsqrt_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+rsub_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+rsub_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+rwkv_ka_fusion:
+  shape_desc: M, N
+  shapes:
+  - [64]
+  - [256]
+  - [1024]
+  - [4096]
+  - [8192]
+  - [16384]
+  - [32768]
+rwkv_mm_sparsity:
+  shape_desc: M, N
+  shapes:
+  - [1]
+safe_softmax:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+scaled_dot_product_attention:
+  shape_desc: M, N
+  shapes:
+  - [4, 32, 1024, 64]
+  - [4, 32, 1024, 128]
+  - [4, 32, 2048, 128]
+  - [4, 32, 4096, 128]
+  - [4, 32, 8192, 128]
+scaled_mm:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 16, 16]
+  - [128, 128, 128]
+  - [512, 512, 512]
+scaled_mm_out:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 16, 16]
+  - [128, 128, 128]
+  - [512, 512, 512]
+scaled_softmax_backward:
+  shape_desc: batch, heads, query_len, key_len
+  shapes:
+  - [1, 4, 64, 64]
+  - [2, 8, 128, 128]
+  - [4, 8, 256, 256]
+  - [8, 12, 512, 512]
+  - [16, 16, 1024, 1024]
+  - [32, 16, 512, 512]
+  - [1, 32, 2048, 2048]
+  - [2, 40, 4096, 4096]
+scaled_softmax_forward:
+  shape_desc: batch, heads, query_len, key_len
+  shapes:
+  - [1, 4, 64, 64]
+  - [2, 8, 128, 128]
+  - [4, 8, 256, 256]
+  - [8, 12, 512, 512]
+  - [16, 16, 1024, 1024]
+  - [32, 16, 512, 512]
+  - [1, 32, 2048, 2048]
+  - [2, 40, 4096, 4096]
+scatter_add_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+scatter_reduce.two:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+scatter_src:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+scatter_src_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+searchsorted:
+  shape_desc: sorted_sequence shape, values shape, sorter, right, out_int32
+  shapes:
+  - - [1024]
+    - [4096]
+    - false
+    - false
+    - false
+  - - [64, 256]
+    - [64, 256]
+    - false
+    - true
+    - false
+  - - [256, 1024]
+    - [256, 512]
+    - false
+    - false
+    - true
+searchsorted_out:
+  shape_desc: sorted_sequence shape, values shape, sorter, right, out_int32
+  shapes:
+  - - [1024]
+    - [4096]
+    - false
+    - false
+    - false
+  - - [64, 256]
+    - [64, 256]
+    - false
+    - true
+    - false
+  - - [256, 1024]
+    - [256, 512]
+    - false
+    - false
+    - true
+searchsorted_scalar:
+  shape_desc: sorted_sequence shape, right, out_int32
+  shapes:
+  - - [1024]
+    - false
+    - false
+  - - [4096]
+    - true
+    - false
+  - - [8192]
+    - false
+    - true
+searchsorted_scalar_out:
+  shape_desc: sorted_sequence shape, right, out_int32
+  shapes:
+  - - [1024]
+    - false
+    - false
+  - - [4096]
+    - true
+    - false
+  - - [8192]
+    - false
+    - true
+segment_reduce:
+  shape_desc: data shape
+  shapes: &id001
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+segment_reduce_backward:
+  shape_desc: data shape
+  shapes: *id001
+segment_reduce_backward_out:
+  shape_desc: data shape
+  shapes: *id001
+segment_reduce_out:
+  shape_desc: data shape
+  shapes: *id001
+select_backward:
+  shape_desc: M, N
+  shapes:
+  - [128, 256]
+  - [1024, 1024]
+  - [512, 1024, 512]
+  - [16, 8192, 4096]
+  - [8, 4096, 11008]
+  - [4, 32, 4096, 128]
+  - [32, 256, 256, 128]
+select_scatter:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+selu:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+selu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sgn_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sigmoid:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sigmoid_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sigmoid_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+signbit:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+silu:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+silu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+silu_and_mul:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+silu_and_mul_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+silu_and_mul_with_clamp:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+silu_and_mul_with_clamp_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+silu_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sin:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sin_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+single_dim_shapes:
+  shape_desc: M
+  shapes:
+  - [64]
+  - [1024]
+  - [2048]
+  - [4096]
+  - [65536]
+sinh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+skip_layer_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+slice_backward:
+  shape_desc: M, N
+  shapes:
+  - [128, 256]
+  - [1024, 1024]
+  - [512, 1024, 512]
+  - [16, 8192, 4096]
+  - [8, 4096, 11008]
+  - [4, 32, 4096, 128]
+  - [32, 256, 256, 128]
+slice_scatter:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 256]
+  - [10000, 65536]
+smooth_l1_loss:
+  shape_desc: (B), M, N
+  shapes:
+  - [262144]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+smooth_l1_loss_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [262144]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+soft_margin_loss:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+softmax:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+softmax_backward:
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [256, 1024, 1024]
+softmax_backward_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+softmax_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+softplus:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+softshrink:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+softshrink_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sort:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 512]
+  - [16, 131072]
+  - [8, 262144]
+sort_stable:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [1024, 1]
+  - [1024, 512]
+  - [16, 131072]
+  - [8, 262144]
+sparse_attention:
+  shape_desc: B, M, KV_LEN, TOPK, H, D
+  shapes:
+  - [16, 1, 136, 136, 8, 512]
+  - [16, 1, 392, 385, 8, 512]
+  - [16, 1, 392, 386, 8, 512]
+  - [16, 1, 392, 387, 8, 512]
+  - [32, 1, 392, 388, 8, 512]
+  - [32, 1, 392, 389, 8, 512]
+  - [32, 1, 392, 390, 8, 512]
+  - [32, 1, 392, 391, 8, 512]
+  - [64, 1, 136, 136, 8, 512]
+  - [64, 1, 392, 385, 8, 512]
+  - [64, 1, 392, 388, 8, 512]
+  - [64, 1, 392, 389, 8, 512]
+special_i0e:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+special_i0e_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+special_i1:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+special_i1_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+split_with_sizes_copy:
+  shape_desc: M, N
+  shapes:
+  - [10]
+  - [10, 4]
+  - [10, 4, 8]
+  - [10, 4, 8, 16]
+  - [16, 32]
+  - [8, 64, 128]
+  - [1, 8192]
+  - [32, 50257]
+sqrt:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+sqrt_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+square:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+square_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+square_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+stack:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [64, 64, 1]
+  - [64, 64, 16]
+std:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+sub:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+sub_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+  - [64, 64, 65536]
+sum:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+sum_dim:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+sum_dim_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+sum_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+svd:
+  shape_desc: (*B), M, N
+  shapes:
+  - [4096, 2, 2]
+  - [4096, 16, 2]
+  - [4096, 2, 16]
+  - [4096, 8, 8]
+  - [256, 16, 8]
+  - [256, 8, 16]
+  - [64, 32]
+  - [32, 64]
+  - [16, 16, 16]
+  - [17, 17]
+  - [256, 256]
+swiglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [1024, 8192]
+  - [1024, 131072]
+  - [64, 64, 2]
+  - [64, 64, 32]
+  - [64, 64, 512]
+  - [64, 64, 8192]
+t_copy:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+t_copy_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+tan:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+tan_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+tanh:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+tanh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+tanh_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+tensor_split:
+  shape_desc: M, N
+  shapes:
+  - [64]
+  - [256]
+  - [1024]
+  - [64, 64]
+  - [128, 128]
+  - [256, 256]
+  - [512, 512]
+  - [8, 16, 32]
+  - [16, 32, 64]
+  - [32, 64, 128]
+  - [4, 8, 16, 32]
+  - [8, 16, 32, 64]
+threshold:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+threshold_backward:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+tile:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [16, 256, 256]
+  - [512, 512, 512]
+  - [64, 64, 64, 64]
+top_k_per_row_decode:
+  shape_desc: vocab_size, top_k
+  shapes:
+  - [129280, 1024]
+  - [32768, 512]
+  - [16384, 256]
+  - [8192, 128]
+  - [4096, 64]
+top_k_per_row_prefill:
+  shape_desc: num_rows, vocab_size, top_k
+  shapes:
+  - [1, 129280, 1024]
+  - [32, 129280, 1024]
+  - [64, 129280, 1024]
+  - [2048, 129280, 1024]
+topk:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [4096, 4096]
+  - [10000, 256]
+  - [10000, 65536]
+  - [4, 128]
+  - [8, 256]
+  - [64, 128, 8]
+  - [64, 1024, 32]
+  - [64, 8192, 128]
+  - [128, 32768, 256]
+  - - [4, 128, 64]
+    - 5
+  - - [4, 128, 64]
+    - 64
+  - - [8, 512, 32]
+    - 32
+  - - [16, 1024, 256]
+    - 256
+topk_softmax:
+  shape_desc: M, N
+  shapes:
+  - [256, 128, 16]
+  - [1024, 256, 32]
+  - [4096, 64, 8]
+  - [8192, 128, 8]
+  - [16384, 256, 8]
+topk_softplus_sqrt:
+  shape_desc: num_tokens, num_experts, topk
+  shapes:
+  - [1, 256, 6]
+  - [10, 256, 6]
+  - [16, 256, 6]
+  - [128, 256, 6]
+  - [512, 256, 6]
+  - [1024, 256, 6]
+  - [2048, 256, 6]
+  - [4096, 256, 6]
+tril:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 16, 16]
+  - [512, 32, 32]
+  - [128, 64, 64]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+tril_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+tril_extreme_diagonal:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+tril_out:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+tril_out_sliced:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 1024]
+  - [16, 128, 128]
+  - [512, 32, 32]
+tril_out_transposed:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 1024]
+  - [16, 128, 128]
+  - [512, 32, 32]
+triu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+triu_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+trunc_divide:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+trunc_divide_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+unfold_backward:
+  shape_desc: M, N
+  shapes:
+  - - [32, 64]
+    - 1
+    - 16
+    - 16
+  - - [16, 33]
+    - 0
+    - 5
+    - 2
+  - - [4, 8, 12]
+    - -1
+    - 6
+    - 4
+  - - [7, 13]
+    - 1
+    - 13
+    - 3
+  - - [6, 20]
+    - 1
+    - 7
+    - 4
+  - - [2, 3, 17]
+    - -1
+    - 9
+    - 1
+  - - [2, 17]
+    - 1
+    - 4
+    - 6
+uniform_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+unique2:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+unique_consecutive:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+unpack_seq_triton:
+  shape_desc: N, D, B, lengths
+  shapes:
+  - - 512
+    - 64
+    - 5
+    - [64, 128, 64, 128, 128]
+  - - 4096
+    - 128
+    - 4
+    - [1024, 1024, 1024, 1024]
+  - - 8192
+    - 256
+    - 5
+    - [1024, 2048, 1024, 2048, 2048]
+  - - 2048
+    - 512
+    - 4
+    - [512, 512, 512, 512]
+  - - 16384
+    - 64
+    - 8
+    - [2048, 2048, 2048, 2048, 2048, 2048, 2048, 2048]
+  - - 1024
+    - 1024
+    - 4
+    - [256, 256, 256, 256]
+upsample_bicubic2d_aa:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+upsample_bicubic2d_aa_backward:
+  shape_desc: N, C, input_h, input_w, output_h, output_w, align_corners
+  shapes:
+  - [4, 16, 4, 4, 1, 1, false]
+  - [4, 16, 4, 4, 16, 16, false]
+  - [4, 16, 16, 16, 4, 4, false]
+  - [4, 16, 16, 32, 64, 128, false]
+  - [1, 1, 64, 64, 16, 16, false]
+  - [1, 1, 64, 64, 32, 32, false]
+  - [1, 1, 64, 64, 128, 128, false]
+  - [4, 3, 256, 256, 128, 128, false]
+  - [4, 3, 128, 128, 256, 256, false]
+  - [4, 64, 64, 64, 32, 32, false]
+  - [1, 64, 512, 512, 128, 128, false]
+  - [1, 64, 512, 512, 1024, 1024, false]
+  - [512, 1024, 32, 32, 8, 8, false]
+  - [256, 512, 64, 64, 16, 16, false]
+  - [256, 512, 64, 64, 32, 32, false]
+  - [256, 512, 64, 64, 128, 128, false]
+upsample_linear1d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+upsample_nearest1d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+upsample_nearest2d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+upsample_nearest3d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+upsample_nearest_exact1d:
+  shape_desc: M, N
+  shapes:
+  - [2, 3, 16]
+  - [4, 8, 64]
+  - [8, 16, 256]
+  - [16, 32, 512]
+upsample_trilinear3d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+var:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+var_correction:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+var_mean:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1049600]
+  - [1073741824]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [1024, 1048576]
+  - [64, 1, 64]
+  - [64, 16, 64]
+  - [64, 256, 64]
+  - [64, 4096, 64]
+vdot:
+  shape_desc: M
+  shapes:
+  - [64]
+  - [1024]
+  - [2048]
+  - [4096]
+  - [65536]
+vector_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+view_copy:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 1]
+  - [1024, 16]
+  - [1024, 256]
+  - [1024, 4096]
+  - [1024, 65536]
+  - [64, 64, 1]
+  - [64, 64, 16]
+  - [64, 64, 256]
+  - [64, 64, 4096]
+vstack:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [1024, 2]
+  - [1024, 32]
+  - [1024, 512]
+  - [64, 64, 1]
+  - [64, 64, 16]
+w8a8_block_fp8_matmul:
+  shape_desc: M, N, K
+  shapes:
+  - [64, 128, 128]
+  - [128, 256, 512]
+  - [1, 4096, 7168]
+  - [16, 4096, 7168]
+  - [64, 4096, 7168]
+  - [83, 7748, 3884]
+  - [84, 7168, 3884]
+w8a8_block_fp8_matmul_deepgemm:
+  shape_desc: M, N, K
+  shapes:
+  - [64, 128, 128]
+  - [128, 256, 512]
+  - [1, 4096, 7168]
+  - [16, 4096, 7168]
+  - [64, 4096, 7168]
+weight_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+weight_norm_interface:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+where_self:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+where_self_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+zero:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+zero_:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+zero_out:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+zeros:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
+zeros_like:
+  shape_desc: M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+  - [268435456]
+  - [10000, 1]
+  - [10000, 256]
+  - [10000, 65536]
+  - [100, 1, 100]
+  - [100, 256, 100]
+  - [100, 65536, 100]
diff --git a/ops/shapes_single.yaml b/ops/shapes_single.yaml
new file mode 100644
index 0000000..c531289
--- /dev/null
+++ b/ops/shapes_single.yaml
@@ -0,0 +1,2712 @@
+# 由 gen_inventory.py 从 shapes_single.yaml 迁移：保留新上游仍存在的 op_name 键 555/601，并合并上游 core_shapes.yaml 缺失键（main:+43 vllm:+3）。
+# 顶层键 = FlagGems op_name 或 Benchmark 类名（上游回退链需要）；
+# 两者都无键的算子使用其基类 DEFAULT_SHAPES。
+AttentionBenchmark:
+  shapes:
+  - [4, 32, 1024, 64]
+  - [4, 32, 1024, 128]
+  - [4, 32, 2048, 128]
+  - [4, 32, 4096, 128]
+  - [4, 32, 8192, 128]
+Benchmark:
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+BlasBenchmark:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+  - [2, 4096, 4096, 4096]
+  - [16, 1024, 1024, 1024]
+  - [16, 2048, 2048, 2048]
+  - [16, 4096, 4096, 4096]
+ChunkGatedDeltaRuleBenchmark:
+  shape_desc: B, T, Hg, H, K, V
+  shapes:
+  - [1, 64, 4, 8, 64, 64]
+  - [1, 128, 4, 8, 64, 64]
+  - [1, 256, 4, 8, 64, 64]
+  - [1, 512, 4, 8, 64, 64]
+  - [2, 64, 4, 8, 64, 64]
+  - [2, 128, 4, 8, 64, 64]
+  - [2, 256, 4, 8, 64, 64]
+  - [2, 512, 4, 8, 64, 64]
+ConcatAndCacheMLABenchmark:
+  shapes:
+  - [16, 8, 84, 32, 4]
+  - [32, 4, 21, 16, 16]
+  - [64, 8, 21, 8, 4]
+  - [512, 64, 42, 16, 8]
+  - [1024, 64, 84, 16, 8]
+ConcatBenchmark:
+  shape_desc: ((B), M, N) * 3
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [512, 512]
+  - [512, 1024]
+  - [512, 2048]
+Conv1DBenchmark:
+  shape_desc: N, C, L, Co, K, stride, padding, groups
+  shapes:
+  - [32, 64, 256, 64, 3, 1, 1, 1]
+  - [64, 32, 512, 32, 5, 2, 2, 1]
+  - [16, 16, 1024, 32, 7, 1, 3, 1]
+  - [32, 32, 4096, 64, 9, 2, 4, 2]
+Conv2DBenchmark:
+  shape_desc: N, C_in, H, W, C_out, K_h, K_w, stride, padding, groups
+  shapes:
+  - [32, 64, 128, 128, 32, 3, 3, 1, 1, 1]
+  - [104, 16, 32, 32, 32, 4, 4, 1, 0, 1]
+  - [64, 32, 18, 18, 32, 5, 5, 2, 1, 1]
+  - [64, 32, 210, 210, 64, 5, 5, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 3, 3, 1, 1, 2]
+Conv3DBenchmark:
+  shape_desc: N, C_in, D, H, W, C_out, K_d, K_h, K_w, stride, padding, groups
+  shapes:
+  - [32, 64, 64, 64, 64, 32, 3, 3, 3, 1, 1, 1]
+  - [32, 64, 64, 64, 64, 32, 3, 3, 3, 2, 1, 1]
+  - [16, 32, 24, 24, 24, 24, 3, 3, 3, 1, 1, 2]
+  - [64, 32, 18, 18, 18, 32, 5, 5, 5, 1, 2, 1]
+  - [16, 32, 12, 12, 12, 24, 3, 3, 3, 1, 2, 1]
+ConvDepthwise2DBenchmark:
+  shape_desc: N, C, H, W, K_h, K_w, stride, padding, dilation
+  shapes:
+  - [32, 32, 64, 64, 3, 3, 1, 1, 1]
+  - [16, 64, 128, 128, 3, 3, 2, 1, 1]
+  - [8, 128, 56, 56, 5, 5, 1, 2, 1]
+  - [4, 256, 28, 28, 3, 3, 1, 1, 2]
+  - [2, 512, 14, 14, 3, 3, 1, 1, 1]
+EmbeddingBenchmark:
+  shapes:
+  - [4, 4]
+  - [16, 16]
+  - [128, 128]
+  - [256, 256]
+  - [1024, 1024]
+FlashMLABenchmark:
+  shapes:
+  - [1024]
+  - [2048]
+  - [4096]
+  - [8192]
+  - [16384]
+FractionalMaxPool2dBenchmark:
+  shape_desc: N, C, H, W
+  shapes:
+  - [4, 3, 224, 224]
+  - [16, 64, 56, 56]
+  - [32, 128, 28, 28]
+  - [64, 256, 14, 14]
+  - [128, 512, 7, 7]
+FusedRecurrentGatedDeltaRuleBenchmark:
+  shape_desc: (T,) many lengths
+  shapes:
+  - [1]
+  - [64]
+  - [184]
+  - [256]
+  - [512]
+GenericBenchmark2DOnly:
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+GenericBenchmark4DOnly:
+  shapes:
+  - [64, 64, 4, 8]
+  - [256, 256, 4, 8]
+  - [1024, 1024, 4, 8]
+  - [4096, 4096, 4, 8]
+  - [1024, 65536, 4, 8]
+GenericBenchmarkExcluse1D:
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+GenericBenchmarkExcluse3D:
+  shapes:
+  - [1048576]
+  - [1073741824]
+  - [1024, 512]
+  - [1024, 1024]
+  - [4096, 4096]
+IndexPutAccFalseBenchmark:
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [2, 8]
+    - [8]
+    - false
+  - - [1024, 1024]
+    - - [4, 64]
+    - [1024]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+    - [512]
+    - false
+IndexPutAccTrueBenchmark:
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+  - - [32, 32]
+    - - [8]
+      - [8]
+    - [8]
+    - false
+  - - [1024, 1024]
+    - - [64]
+      - [64]
+    - [64]
+    - false
+  - - [512, 512, 512]
+    - - [128]
+      - [128]
+      - [128]
+    - [128]
+    - false
+  - - [512, 512, 512]
+    - - [2, 128]
+      - [2, 128]
+      - [2, 128]
+    - [2, 128]
+    - false
+IndexReduceBenchmark:
+  shape_desc: (B), M, N
+  shapes:
+  - [1048576]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [64, 512, 512]
+KronBenchmark:
+  shapes:
+  - [16, 16]
+  - [64, 64]
+  - [128, 128]
+  - [256, 256]
+  - [4, 8, 16, 32]
+MoeAlignBlockSizeBenchmark:
+  shapes:
+  - [16, 8]
+  - [4, 32]
+  - [32, 16]
+  - [8, 64]
+  - [128, 4]
+MvAndOuterBenchmark:
+  shapes:
+  - [384, 384]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [8192, 8192]
+  - [10240, 10240]
+NormBenchmark:
+  shape_desc: N, C, *
+  shapes:
+  - [4, 16, 64, 4]
+  - [16, 16, 8, 48]
+  - [16, 16, 8, 88]
+  - [16, 16, 128]
+  - [20, 6, 65536]
+PerTokenGroupQuantFp8Benchmark:
+  shapes:
+  - [7, 512, 512]
+  - [7, 4096, 256]
+  - [83, 512, 64]
+  - [2048, 4096, 256]
+  - [2048, 13824, 512]
+RWKVBenchmark:
+  shapes:
+  - [64]
+  - [256]
+  - [1024]
+  - [4096]
+  - [8192]
+  - [16384]
+  - [32768]
+RWKVSparsityBenchmark:
+  shapes:
+  - [1]
+ReshapeAndCacheBenchmark:
+  shapes:
+  - [21, 4, 64, 8, 512]
+  - [42, 8, 64, 8, 1024]
+  - [42, 8, 80, 16, 10000]
+  - [42, 8, 120, 32, 10000]
+  - [42, 8, 256, 8, 1024]
+ReshapeAndCacheFlashBenchmark:
+  shapes:
+  - [21, 4, 64, 8, 512]
+  - [21, 4, 64, 8, 1024]
+  - [21, 4, 80, 16, 10000]
+  - [42, 8, 64, 8, 1024]
+  - [42, 8, 80, 16, 10000]
+RopeBenchmark:
+  shapes:
+  - [128]
+  - [192]
+  - [256]
+  - [384]
+  - [512]
+ScaledDotProductFusedAttentionOverrideableBenchmark:
+  shape_desc: B, H, S, D
+  shapes:
+  - [2, 8, 64, 64]
+  - [4, 8, 128, 64]
+  - [2, 8, 128, 128]
+  - [1, 16, 256, 64]
+TensorRepeatBenchmark:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [512, 512]
+  - [512, 1024]
+  - [512, 2048]
+TensorSelectBenchmark:
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+UnaryPointwiseBenchmark:
+  shape_desc: (B), M, N
+  shapes:
+  - [1073741824]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+UnaryReductionBenchmark:
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+UpsampleBenchmark:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+  - [8, 16, 128, 128]
+  - [2, 3, 1024, 1024]
+  - [16, 16, 512, 512]
+  - [16, 16, 1024, 1024]
+abs:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+abs_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+absolute:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+acos:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+act_quant_triton:
+  shape_desc: M, N
+  shapes:
+  - [1, 128]
+add_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+add_rms_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+addcdiv:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+addcdiv_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+addcmul:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+addcmul_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+addmm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 2048, 2048, 2048]
+addmm_dtype:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 2048, 2048, 2048]
+addmm_dtype_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 2048, 2048, 2048]
+addmm_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 2048, 2048, 2048]
+addmv:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+addr:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 1024, 1024, 1024]
+alias_copy:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+alias_copy_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+all:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+all_dim:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+allclose:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+amax:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+aminmax:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+angle:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+any:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+any_dim:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+any_dims:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+apply_repetition_penalties:
+  shape_desc: M, N
+  shapes:
+  - [64, 32000]
+apply_rotary_pos_emb:
+  shape_desc: M, N
+  shapes:
+  - [512]
+arange:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+arange_start:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+arange_start_step:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+arcsinh:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+arcsinh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+arcsinh_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+arctanh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+argmax:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+argmin:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+argsort:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+as_strided_copy:
+  shape_desc: input shape, size, stride, storage_offset
+  shapes:
+  - - [64, 64]
+    - [64, 64]
+    - [64, 1]
+    - 0
+as_strided_copy_out:
+  shape_desc: input shape, size, stride, storage_offset
+  shapes:
+  - - [64, 64]
+    - [64, 64]
+    - [64, 1]
+    - 0
+asinh:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+asinh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+assert_async:
+  shape_desc: M, N
+  shapes:
+  - []
+atan:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+atan2:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+atan2_out:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+atan_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+atanh:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+avg_pool2d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [32, 128, 28, 28]
+avg_pool2d_backward:
+  shape_desc: N, C, H, W
+  shapes:
+  - [32, 128, 28, 28]
+avg_pool3d:
+  shape_desc: N, C, D, H, W
+  shapes:
+  - [8, 64, 8, 28, 28]
+baddbmm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 1024, 1024, 1024]
+batch_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [20, 6, 65536]
+batch_norm_backward:
+  shape_desc: N, C, *
+  shapes:
+  - [20, 6, 65536]
+bernoulli_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bincount:
+  shape_desc: M, N
+  shapes:
+  - [10000, 1]
+bincount_weighted:
+  shape_desc: M, N
+  shapes:
+  - [10000, 1]
+bitwise_and_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_and_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_and_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_and_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_and_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_left_shift:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_not:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+bitwise_not_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+bitwise_or_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_or_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_or_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_or_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_or_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bitwise_right_shift:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+bmm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 1024, 1024, 1024]
+bmm_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [16, 1024, 1024, 1024]
+cat_out:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+cauchy:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+cauchy_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+cdist_backward:
+  shape_desc: batch, n1, dim
+  shapes:
+  - [2, 16, 32]
+  - [4, 32, 64]
+  - [8, 64, 128]
+ceil:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+ceil_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+ceil_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+celu:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+celu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+chunk_gated_delta_rule_fwd:
+  shape_desc: B, T, H, K, V
+  shapes:
+  - [2, 16384, 16, 128, 128]
+  - [4, 2048, 16, 128, 128]
+  - [4, 4096, 64, 128, 128]
+chunk_gdn2:
+  shapes:
+  - [2, 512, 8, 64, 64]
+  - [4, 1024, 8, 64, 64]
+  - [1, 2048, 8, 64, 64]
+  - [1, 4096, 16, 64, 64]
+  - [1, 8192, 96, 128, 128]
+  - [2, 2048, 16, 256, 512]
+  - [2, 16384, 16, 128, 128]
+  - [4, 1024, 8, 256, 512]
+  - [4, 2048, 16, 128, 128]
+  - [4, 4096, 64, 128, 128]
+  - [8, 1024, 8, 64, 64]
+  - [8, 2048, 32, 256, 256]
+clamp:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clamp_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clamp_max:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+clamp_max_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+clamp_min:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clamp_min_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clamp_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clamp_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clip:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+clip_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+col2im:
+  shape_desc: M, N
+  shapes:
+  - - 4
+    - 32
+    - [3, 3]
+    - [32, 32]
+    - [1, 1]
+    - [1, 1]
+    - [1, 1]
+combine_topk_swa_indices:
+  shape_desc: M, N
+  shapes:
+  - - [3, 2]
+    - [6, 4]
+    - [4, 3]
+    - 4
+    - 4
+    - 2
+    - 20
+    - 8
+compute_global_topk_indices_and_lens:
+  shape_desc: M, N
+  shapes:
+  - [5, 4, 2, 4, 64]
+concat_and_cache_mla:
+  shape_desc: M, N
+  shapes:
+  - [16, 8, 84, 32, 4]
+concatenate:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+conj_physical:
+  shape_desc: M, N
+  shapes:
+  - [2048, 2048]
+constant_pad_nd:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+contiguous:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+conv1d:
+  shape_desc: N, C, L, Co, K, stride, padding, groups
+  shapes:
+  - [64, 32, 512, 32, 5, 2, 2, 1]
+conv1d_padding:
+  shape_desc: batch, input_c, input_l, out_c, kernel, stride, padding, groups
+  shapes:
+  - [64, 48, 1024, 128, 5, 2, 2, 1]
+conv2d:
+  shape_desc: N, C_in, H, W, C_out, K_h, K_w, stride, padding, groups
+  shapes:
+  - [104, 16, 32, 32, 32, 4, 4, 1, 0, 1]
+conv2d_padding:
+  shape_desc: batch, input_c, input_h, input_w, out_c, kernel_h, kernel_w, stride,
+    padding, groups
+  shapes:
+  - [32, 64, 128, 128, 32, 3, 3, 1, same, 1]
+conv3d:
+  shape_desc: N, C_in, D, H, W, C_out, K_d, K_h, K_w, stride, padding, groups
+  shapes:
+  - [16, 32, 24, 24, 24, 24, 3, 3, 3, 1, 1, 2]
+conv3d_padding:
+  shape_desc: batch, input_c, input_d, input_h, input_w, out_c, kernel_d, kernel_h,
+    kernel_w, stride, padding, groups
+  shapes:
+  - [32, 8, 8, 8, 8, 32, 3, 3, 3, 1, valid, 1]
+conv_depthwise2d:
+  shape_desc: N, C, H, W, K_h, K_w, stride, padding, dilation
+  shapes:
+  - [32, 32, 64, 64, 3, 3, 1, 1, 1]
+conv_transpose1d:
+  shape_desc: batch, input_c, input_l, out_c, kernel, stride, padding, groups
+  shapes:
+  - [64, 48, 256, 128, 5, 2, 2, 1]
+conv_transpose2d:
+  shape_desc: N, C_in, H, W, C_out, K_h, K_w, stride, padding, groups
+  shapes:
+  - [16, 32, 32, 32, 64, 3, 3, 2, 1, 1]
+copy:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+copy_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+copysign:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+copysign_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+cos:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+cos_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+cosh:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 65536]
+cosh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 65536]
+cosh_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 65536]
+count_nonzero:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+cp_gather_indexer_k_quant_cache:
+  shape_desc: batch_size, seq_len, block_size, head_dim, quant_block_size
+  shapes:
+  - [32, 1024, 16, 512, 128]
+cross_entropy_loss:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+ctc_loss:
+  shape_desc: T, N, C, S
+  shapes:
+  - [256, 16, 64, 48]
+ctc_loss_backward:
+  shape_desc: T, N, C, S
+  shapes:
+  - [64, 4, 32, 16]
+  - [256, 16, 64, 48]
+  - [512, 32, 64, 48]
+  - [1024, 32, 128, 96]
+cudnn_convolution:
+  shape_desc: batch, input_c, input_h, input_w, out_c, kernel_h, kernel_w, stride,
+    padding, groups
+  shapes:
+  - [16, 32, 12, 12, 24, 3, 3, 2, 1, 1]
+cummax:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+cummin:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+cumprod:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+cumprod_:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+cumsum:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+cutlass_scaled_mm:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 24576, 4096]
+dequantize_and_gather_k_cache:
+  shape_desc: M, N
+  shapes:
+  - [1, 512, 128, 512, 448, 64]
+dgeglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+diag:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+diag_embed:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+diagonal_backward:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+diff:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+digamma_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+div_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+div_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+dot:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+dreglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+dropout:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+dropout_backward:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+dswiglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+dunder_ior_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+dunder_ior_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+dunder_or_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+dunder_or_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+einsum:
+  shape_desc: B, M, N, K
+  shapes:
+  - [1, 1024, 1024, 1024]
+elu:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+elu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+elu_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+embedding:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+embedding_backward:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+embedding_dense_backward:
+  shape_desc: M, N
+  shapes:
+  - [8, 4096, 256, 32768]
+eq:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+eq_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+erf:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+erf_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+exp:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+exp2:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+exp2_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+exp_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+exp_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+expm1:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+expm1_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+expm1_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+exponential_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+eye:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+eye_m:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+feature_dropout:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+feature_dropout_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fft:
+  shape_desc: M, N
+  shapes:
+  - [4096, 512]
+fill_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fill_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fill_scalar_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fill_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fill_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fill_tensor_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+flash_mla:
+  shape_desc: M, N
+  shapes:
+  - [1024]
+flash_mla_sparse_fwd:
+  shape_desc: s_q, s_kv, topk, h_q, d_qk
+  shapes:
+  - [4096, 8192, 2048, 128, 576]
+flip:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+float_power_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 65536]
+floor:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+floor_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+floor_divide_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+floor_divide_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+floor_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+fmin:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fmin_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fmod_:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+fmod_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fmod_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fmod_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fmod_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fp8_einsum:
+  shape_desc: M, N
+  shapes:
+  - [128, 16, 7168, 1024]
+fp8_paged_mqa_logits_gems_vs_deepgemm:
+  shape_desc: batch_size, next_n, heads, head_dim, avg_kv
+  shapes:
+  - [1, 1, 16, 64, 1024]
+full:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+full_like:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+functional_sym_constrain_range_for_size:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fused_add_rms_norm:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+fused_deepseek_v4_qnorm_rope_kv_rope_quant_insert:
+  shape_desc: num_tokens, num_heads, num_tokens_insert, block_size, max_pos, eps
+  shapes:
+  - [2048, 128, 2048, 64, 4096, 1.0e-06]
+fused_experts_impl:
+  shape_desc: num_tokens, num_experts, hidden_size, intermediate_size, topk
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+fused_inv_rope_fp8_quant:
+  shape_desc: num_tokens, num_heads, n_groups, tma_aligned_scales
+  shapes:
+  - [1024, 128, 8, false]
+fused_marlin_moe:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+fused_marlin_moe_int8:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+fused_moe:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+fused_moe_int8_w8a16_gems_vs_bf16_deq:
+  shape_desc: M, N
+  shapes:
+  - [1, 8, 4096, 14336, 2]
+fused_q_kv_rmsnorm:
+  shape_desc: M, N
+  shapes:
+  - [1, 1536, 512]
+gather:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+gather_backward:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+gcd:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 256, 256]
+gcd_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+ge:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+ge_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+geglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+gelu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+gelu_and_mul:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+gelu_backward:
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [256, 1024, 1024]
+glu:
+  shape_desc: M, N
+  shapes:
+  - [4, 8, 2048, 128]
+glu_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+greater:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+greater_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+greater_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+greater_scalar_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+grid_sample:
+  shape_desc: M, N
+  shapes:
+  - [4, 64, 64, 64]
+group_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [20, 6, 65536]
+grouped_mm:
+  shape_desc: Groups, N, K
+  shapes:
+  - [16, 512, 2048]
+  - [16, 2560, 2048]
+  - [64, 2048, 128]
+grouped_topk:
+  shape_desc: M, N
+  shapes:
+  - [1024, 64, 8, 2, 8]
+groupmm:
+  shape_desc: Groups, N, K
+  shapes:
+  - [16, 512, 2048]
+  - [16, 2560, 2048]
+  - [64, 2048, 128]
+gt:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+gt_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+hadamard_transform:
+  shape_desc: batch, dim
+  shapes:
+  - [64, 64]
+hadamard_transform_mn:
+  shape_desc: batch, dim
+  shapes:
+  - [1024, 40960]
+hardsigmoid:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+hardswish_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+hc_head_fused_kernel:
+  shape_desc: N, hidden_size
+  shapes:
+  - [4096, 4096, 4]
+hc_split_sinkhorn_forward:
+  shape_desc: batch, seqlen
+  shapes:
+  - [256, 256, 4]
+histc:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+hstack:
+  shapes:
+  - [536870912]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [512, 512, 512]
+huber_loss:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [256, 256]
+  - [1024, 1024]
+  - [4096, 4096]
+  - [1024, 65536]
+  - [2048, 32768]
+  - [2048, 65536]
+  - [4096, 32768]
+  - [4096, 65536]
+  - [3072, 49152]
+  - [3072, 65536]
+hypot:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+hypot_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+i0:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+i0_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+i0_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+index:
+  shape_desc: M, N
+  shapes:
+  - - [268435456]
+    - - [65536]
+index_add:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+index_add_:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+index_copy:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+index_copy_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+index_put:
+  shape_desc: input_shape, [index_tensor_shapes], values_shape, accumulate
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+index_put_:
+  shape_desc: input_shape, [index_tensor_shapes], values_shape, accumulate
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+index_put_impl:
+  shape_desc: input_shape, [index_tensor_shapes], values_shape, accumulate
+  shapes:
+  - - [268435456]
+    - - [65536]
+    - [65536]
+    - false
+index_select:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+index_select_backward:
+  shape_desc: N, M
+  shapes:
+  - [16, 64]
+  - [32, 128]
+  - [64, 256]
+  - [16, 16, 64]
+  - [32, 32, 128]
+indexer_k_quant_and_cache:
+  shape_desc: num_tokens, num_blocks, block_size, head_dim, quant_block_size
+  shapes:
+  - [2048, 256, 16, 512, 128]
+inplace_fused_experts:
+  shape_desc: M, N
+  shapes:
+  - [16, 8, 4096, 14336, 2]
+instance_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [20, 6, 65536]
+is_all_true:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+isclose:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+isfinite:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+isin:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+isin_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+isin_tensor_scalar:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+isinf:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+isnan:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+isneginf:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+kron:
+  shape_desc: M, N
+  shapes:
+  - [256, 256]
+layer_norm:
+  shape_desc: N, C, *
+  shapes:
+  - [20, 6, 65536]
+layer_norm_backward:
+  shape_desc: N, C, *
+  shapes:
+  - [20, 6, 65536]
+le:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+le_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+leaky_relu:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+leaky_relu_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+lerp_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+lerp_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+lerp_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+lerp_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+lift_fresh_copy:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+linspace:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+log:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+log10:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 65536]
+log10_:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 65536]
+log10_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 65536]
+log1p:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+log1p_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+log_sigmoid:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+log_softmax:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+log_softmax_backward_data:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+log_softmax_backward_data_out:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+log_softmax_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logaddexp:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logaddexp_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logical_and:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logical_and_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logical_not:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+logical_or:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logical_or_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logical_xor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logit:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+logit_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+logspace:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+logsumexp:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+lt:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+lt_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+margin_ranking_loss:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_fill:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_fill_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_fill_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_fill_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_scatter:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_scatter_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+masked_select:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+max:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+max_dim:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+max_pool2d_backward:
+  shape_desc: N, C, H, W
+  shapes:
+  - [32, 128, 28, 28]
+max_pool2d_with_indices:
+  shape_desc: N, C, H, W
+  shapes:
+  - [32, 128, 28, 28]
+max_pool3d:
+  shape_desc: N, C, D, H, W
+  shapes:
+  - [4, 3, 16, 56, 56]
+max_pool3d_backward:
+  shape_desc: N, C, D, H, W
+  shapes:
+  - [4, 3, 16, 56, 56]
+maximum:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+mean:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+mean_dim:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+median:
+  shape_desc: input shape
+  shapes:
+  - [4096]
+median_dim:
+  shape_desc: input shape or [input shape, dim, keepdim]
+  shapes:
+  - - [32, 384, 128]
+    - 1
+    - false
+mhc_bwd:
+  shape_desc: seqlen, n_stream
+  shapes:
+  - [65536, 4]
+mhc_post:
+  shape_desc: N, H
+  shapes:
+  - [4096, 7168]
+mhc_pre:
+  shape_desc: N, hidden_size
+  shapes:
+  - [8192, 2560]
+min:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+min_dim:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+minimum:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+mm:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+mm_out:
+  shape_desc: B, M, N, K
+  shapes:
+  - [2, 384, 384, 384]
+mm_self_transpose:
+  shape_desc: M, K
+  shapes:
+  - [1024, 1024]
+moe_align_block_size_triton:
+  shape_desc: M, N
+  shapes:
+  - [512, 64, 16384, 10]
+moe_sum:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+mse_loss:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+mul:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+mul_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+multinomial:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+mv:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+nan_to_num:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+nanmedian:
+  shape_desc: N or M, N
+  shapes:
+  - [64]
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+nanmedian_dim:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+nanmedian_dim_values:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+nanmedian_out:
+  shape_desc: N or M, N
+  shapes:
+  - [64]
+  - [64, 64]
+  - [1024, 1024]
+  - [256, 4096]
+  - [16, 131072]
+ne:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+ne_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+neg:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+neg_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+new_full:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+nll_loss2d_backward:
+  shape_desc: M, N
+  shapes:
+  - [256, 256, 4, 8]
+nll_loss2d_forward:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024, 4, 8]
+nll_loss_backward:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+nll_loss_forward:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+nll_loss_nd_backward:
+  shape_desc: N, C, d1, d2, ...
+  shapes:
+  - [64, 64]
+nll_loss_nd_forward:
+  shape_desc: N, C, d1, d2, ...
+  shapes:
+  - [32, 128, 512]
+nonzero:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+nonzero_numpy:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+normal_float_float_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+normal_float_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+normal_tensor_float:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+normal_tensor_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+one_hot:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+ones:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+ones_like:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+outer:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+pack_seq_triton:
+  shape_desc: N, D, B, lengths
+  shapes:
+  - - 512
+    - 64
+    - 5
+    - [64, 128, 64, 128, 128]
+pad:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+per_token_group_quant_fp8:
+  shape_desc: M, N
+  shapes:
+  - [7, 4096, 256]
+pixel_shuffle:
+  shape_desc: N, C, H, W, upscale_factor
+  shapes:
+  - [1, 4, 2, 3, 2]
+  - [2, 9, 4, 4, 3]
+  - [4, 64, 32, 32, 2]
+  - [2, 128, 64, 64, 2]
+  - [1, 64, 16, 16, 4]
+  - [8, 36, 64, 64, 3]
+  - [1, 16, 128, 128, 2]
+pixel_unshuffle:
+  shape_desc: M, N
+  shapes:
+  - - [4, 16, 64, 48]
+    - 4
+poisson:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+polar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+pow_tensor_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+pow_tensor_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+prelu:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+prod:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+quantile:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+rad2deg:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+rad2deg_:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+rand:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+rand_like:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+randint:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+randint_like:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+randn:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+randn_like:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+randperm:
+  shape_desc: M
+  shapes:
+  - [65536]
+reciprocal:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+reciprocal_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+reflection_pad1d:
+  shape_desc: M, N
+  shapes:
+  - - [8, 16, 256]
+    - [8, 8]
+reflection_pad1d_backward:
+  shape_desc: M, N
+  shapes:
+  - [2, 3]
+reflection_pad1d_out:
+  shape_desc: M, N
+  shapes:
+  - - [8, 16, 256]
+    - [8, 8]
+reflection_pad2d:
+  shape_desc: M, N
+  shapes:
+  - - [16, 32, 64, 128]
+    - [1, 1, 1, 1]
+reflection_pad2d_out:
+  shape_desc: M, N
+  shapes:
+  - - [16, 32, 64, 128]
+    - [1, 1, 1, 1]
+reglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+relu:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+relu6:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+relu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+remainder:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+remainder_scalar_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+remainder_scalar_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+remainder_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+remainder_tensor_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+renorm:
+  shape_desc: M, N
+  shapes:
+  - [4, 8]
+repeat:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+repeat_interleave_self_int:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+repeat_interleave_self_tensor:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 256, 256]
+  - [128, 256, 256]
+  - [64, 64, 64, 64]
+repeat_interleave_tensor:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 256, 256]
+  - [128, 256, 256]
+  - [64, 64, 64, 64]
+replication_pad1d:
+  shape_desc: M, N
+  shapes:
+  - - [8, 32, 256]
+    - [1, 2]
+replication_pad3d:
+  shape_desc: M, N
+  shapes:
+  - [2, 32, 16, 128, 128]
+reshape_and_cache:
+  shape_desc: M, N
+  shapes:
+  - [42, 8, 80, 16, 10000]
+reshape_and_cache_flash:
+  shape_desc: M, N
+  shapes:
+  - [42, 8, 80, 16, 10000]
+resolve_conj:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+resolve_neg:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+rms_norm:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+roll:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+rot90:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+round:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+round_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+round_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+router_gemm:
+  shape_desc: M, N, K
+  shapes:
+  - [1024, 8, 4096]
+rrelu_with_noise_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+rsqrt:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+rsqrt_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+rsub_scalar:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+rsub_tensor:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+rwkv_ka_fusion:
+  shape_desc: M, N
+  shapes:
+  - [32768]
+rwkv_mm_sparsity:
+  shape_desc: M, N
+  shapes:
+  - [1]
+safe_softmax:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+scaled_dot_product_attention:
+  shape_desc: M, N
+  shapes:
+  - [4, 32, 1024, 64]
+scaled_mm:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 16, 16]
+scaled_mm_out:
+  shape_desc: M, N, K
+  shapes:
+  - [16, 16, 16]
+scaled_softmax_backward:
+  shape_desc: batch, heads, query_len, key_len
+  shapes:
+  - [2, 8, 128, 128]
+scaled_softmax_forward:
+  shape_desc: batch, heads, query_len, key_len
+  shapes:
+  - [2, 8, 128, 128]
+scatter_add_:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+scatter_reduce.two:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+scatter_src:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+scatter_src_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+searchsorted:
+  shape_desc: sorted_sequence shape, values shape, sorter, right, out_int32
+  shapes:
+  - - [1024]
+    - [4096]
+    - false
+    - false
+    - false
+searchsorted_out:
+  shape_desc: sorted_sequence shape, values shape, sorter, right, out_int32
+  shapes:
+  - - [1024]
+    - [4096]
+    - false
+    - false
+    - false
+searchsorted_scalar:
+  shape_desc: sorted_sequence shape, right, out_int32
+  shapes:
+  - - [1024]
+    - false
+    - false
+searchsorted_scalar_out:
+  shape_desc: sorted_sequence shape, right, out_int32
+  shapes:
+  - - [1024]
+    - false
+    - false
+segment_reduce:
+  shape_desc: data shape
+  shapes: &id001
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [1024, 1024, 1024]
+segment_reduce_backward:
+  shape_desc: data shape
+  shapes: *id001
+segment_reduce_backward_out:
+  shape_desc: data shape
+  shapes: *id001
+segment_reduce_out:
+  shape_desc: data shape
+  shapes: *id001
+select_backward:
+  shape_desc: M, N
+  shapes:
+  - [128, 256]
+select_scatter:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+selu:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+selu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sgn_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sigmoid:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sigmoid_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sigmoid_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+signbit:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+silu:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+silu_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+silu_and_mul:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+silu_and_mul_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+silu_and_mul_with_clamp:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+silu_and_mul_with_clamp_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+silu_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sin:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sin_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+single_dim_shapes:
+  shape_desc: M
+  shapes:
+  - [64]
+  - [1024]
+  - [2048]
+  - [4096]
+  - [65536]
+sinh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+skip_layer_norm:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+slice_backward:
+  shape_desc: M, N
+  shapes:
+  - [128, 256]
+slice_scatter:
+  shape_desc: M, N
+  shapes:
+  - [1024, 65536]
+smooth_l1_loss:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+smooth_l1_loss_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+soft_margin_loss:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+softmax:
+  shape_desc: M, N
+  shapes:
+  - [1048576]
+softmax_backward:
+  shapes:
+  - [1048576]
+  - [64, 64]
+  - [4096, 4096]
+  - [64, 512, 512]
+  - [256, 1024, 1024]
+softmax_backward_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+softmax_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+softplus:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+softshrink:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+softshrink_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sort:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+sort_stable:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+sparse_attention:
+  shape_desc: B, M, KV_LEN, TOPK, H, D
+  shapes:
+  - [16, 1, 136, 136, 8, 512]
+special_i0e:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+special_i0e_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+special_i1:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+special_i1_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+split_with_sizes_copy:
+  shape_desc: M, N
+  shapes:
+  - [10, 4]
+sqrt:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+sqrt_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+square:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+square_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+square_out:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+stack:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+std:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+sub:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+sub_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+sum:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+sum_dim:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+sum_dim_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+sum_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+svd:
+  shape_desc: (*B), M, N
+  shapes:
+  - [4096, 8, 8]
+swiglu:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+t_copy:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+t_copy_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+tan:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+tan_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+tanh:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+tanh_:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+tanh_backward:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+tensor_split:
+  shape_desc: M, N
+  shapes:
+  - [64, 64]
+threshold:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+threshold_backward:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+tile:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+top_k_per_row_decode:
+  shape_desc: vocab_size, top_k
+  shapes:
+  - [32768, 512]
+top_k_per_row_prefill:
+  shape_desc: num_rows, vocab_size, top_k
+  shapes:
+  - [1, 129280, 1024]
+topk:
+  shape_desc: M, N
+  shapes:
+  - [128, 32768, 256]
+topk_softmax:
+  shape_desc: M, N
+  shapes:
+  - [8192, 128, 8]
+topk_softplus_sqrt:
+  shape_desc: num_tokens, num_experts, topk
+  shapes:
+  - [4096, 256, 6]
+tril:
+  shape_desc: (B), M, N
+  shapes:
+  - [4096, 4096]
+tril_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+tril_extreme_diagonal:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+tril_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+tril_out_sliced:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 1024]
+  - [16, 128, 128]
+  - [512, 32, 32]
+tril_out_transposed:
+  shape_desc: (B), M, N
+  shapes:
+  - [1024, 1024]
+  - [16, 128, 128]
+  - [512, 32, 32]
+triu:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+triu_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+trunc_divide:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+trunc_divide_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+unfold_backward:
+  shape_desc: M, N
+  shapes:
+  - - [32, 64]
+    - 1
+    - 16
+    - 16
+uniform_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+unique2:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+unique_consecutive:
+  shape_desc: M, N
+  shapes:
+  - [1024, 1024]
+unpack_seq_triton:
+  shape_desc: N, D, B, lengths
+  shapes:
+  - - 512
+    - 64
+    - 5
+    - [64, 128, 64, 128, 128]
+upsample_bicubic2d_aa:
+  shape_desc: N, C, H, W
+  shapes:
+  - [8, 16, 128, 128]
+upsample_bicubic2d_aa_backward:
+  shape_desc: N, C, input_h, input_w, output_h, output_w, align_corners
+  shapes:
+  - [4, 64, 64, 64, 32, 32, false]
+upsample_linear1d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [8, 16, 128, 128]
+upsample_nearest1d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [16, 16, 512, 512]
+upsample_nearest2d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [16, 16, 512, 512]
+upsample_nearest3d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [16, 16, 512, 512]
+upsample_nearest_exact1d:
+  shape_desc: M, N
+  shapes:
+  - [16, 32, 512]
+upsample_trilinear3d:
+  shape_desc: N, C, H, W
+  shapes:
+  - [1, 3, 512, 512]
+var:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+var_correction:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+var_mean:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+vdot:
+  shape_desc: M
+  shapes:
+  - [65536]
+vector_norm:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+view_copy:
+  shape_desc: (B), M, N
+  shapes:
+  - [64, 64]
+vstack:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+w8a8_block_fp8_matmul:
+  shape_desc: M, N, K
+  shapes:
+  - [64, 4096, 7168]
+w8a8_block_fp8_matmul_deepgemm:
+  shape_desc: M, N, K
+  shapes:
+  - [64, 4096, 7168]
+weight_norm:
+  shape_desc: M, N
+  shapes:
+  - [64, 512, 512]
+weight_norm_interface:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+where_self:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+where_self_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+zero:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+zero_:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+zero_out:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+zeros:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
+zeros_like:
+  shape_desc: M, N
+  shapes:
+  - [4096, 4096]
diff --git a/run_batch.py b/run_batch.py
new file mode 100644
index 0000000..716a305
--- /dev/null
+++ b/run_batch.py
@@ -0,0 +1,511 @@
+#!/usr/bin/env python3
+"""多算子批量测试驱动：按算子清单逐个调 run_pytest.sh，多卡分片、串行测量。
+
+每个算子一个独立子进程（CUDA crash 只废单个算子）、独立产物目录
+<batch>/<repo>/<op>/（与单算子跑 run_pytest.sh 的产物结构完全一致，含
+run.log / shapes.yaml / autotune_records / ttgir）。批量层只做编排：
+
+- 分片：按算子名稳定哈希到 GPU（每卡内部串行，测量口径与单算子一致；
+  跨卡并行测量存在功耗耦合噪声，本模式定位为筛查口径，可疑算子回
+  单算子串行流程确认）；
+- 超时：--op-timeout 杀死挂死算子（记 TIMEOUT，不重试，继续跑）；
+- 重试：pytest rc=1（用例失败/flaky）与信号杀（rc>=128）最多重试 2 次，
+  连续两次同 rc 视为确定性失败提前止损；--dtypes 下两级降级——遇上游
+  "can't be supported by this op" 即去掉限制重跑，其余失败且无成功行时
+  也去掉限制最后救一次（dtype_rescue）；
+- 断点续跑：--resume 复用带 .complete 标记的算子（PASS/SKIP 才打标记，
+  FAIL/TIMEOUT 不打、下次重跑）；
+- replay：--replay-root 指向上一次批量目录，逐算子传 REPLAY_FROM
+  <root>/<repo>/<op>，语义与单算子 REPLAY_FROM 相同。
+
+产物：<batch>/ops_status.csv（每算子一行）+ summary.csv（每表行一行，
+由 batch_summary.py 生成）。
+
+示例：
+  # 全量筛查（两仓库全部可收集算子，排除 marked_skip 与 dismiss）
+  python run_batch.py --all --gpus 0,1,2,3,4,5,6,7 --op-timeout 1800
+
+  # 指定算子子集（repo 前缀默认 main；重名用 @文件名 消歧）
+  python run_batch.py --ops softmax vllm:fused_marlin_moe nextafter_@nextafter_
+
+  # 锁 config 复测（口径同单算子 REPLAY_FROM）
+  python run_batch.py --ops-file my_ops.txt --replay-root runs/batch_xxx
+"""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import hashlib
+import os
+import signal
+import subprocess
+import sys
+import threading
+import time
+from collections import Counter
+from concurrent.futures import ThreadPoolExecutor
+from datetime import datetime
+from pathlib import Path
+from typing import Dict, List, Optional, Set, Tuple
+
+import batch_summary
+
+ZL_BENCH = Path(__file__).resolve().parent
+OPS_DIR = ZL_BENCH / "ops"
+LOG = "[batch]"
+MAX_RETRIES = 2
+
+REPO_DIRS = {
+    "main": "/workspace/dev/FlagGems",
+    "vllm": "/workspace/dev/FlagGems-vllm",
+}
+
+STATUS_FIELDS = [
+    "repo", "op", "op_file", "run_name", "gpu", "status", "return_code",
+    "elapsed_ms", "rows_total", "rows_success", "notes", "out_dir",
+]
+
+DTYPE_UNSUPPORTED_MARK = "can't be supported by this op"
+
+
+# ---------------- 清单与选择 ----------------
+
+class Case:
+    __slots__ = ("repo", "op", "op_file", "run_name")
+
+    def __init__(self, repo: str, op: str, op_file: str, run_name: str):
+        self.repo, self.op, self.op_file, self.run_name = repo, op, op_file, run_name
+
+    @property
+    def key(self) -> str:
+        return f"{self.repo}:{self.op}@{self.op_file}"
+
+
+def load_inventory() -> Dict[str, List[Dict[str, str]]]:
+    inv: Dict[str, List[Dict[str, str]]] = {}
+    for label in REPO_DIRS:
+        path = OPS_DIR / f"inventory_{label}.csv"
+        if path.is_file():
+            with path.open(newline="") as f:
+                inv[label] = list(csv.DictReader(f))
+    if not inv:
+        raise SystemExit(f"{LOG} ops/inventory_*.csv 不存在，先跑 ops/gen_inventory.py")
+    return inv
+
+
+def _run_name(row: Dict[str, str], dup_ops: Set[Tuple[str, str]]) -> str:
+    # 目录名默认用 op；仓库内重名时追加 @op_file 消歧（按 inventory 全局判定，
+    # 与选择子集无关，保证 resume/replay 路径稳定）。op_file 可能带子目录
+    # （test_FLA/ 等），做目录名时把 "/" 压成 "_"。
+    op = row["op"]
+    if (row["repo"], op) not in dup_ops:
+        return op
+    return f"{op}@{row['op_file'].replace('/', '_')}"
+
+
+def _dup_ops(inv: Dict[str, List[Dict[str, str]]]) -> Set[Tuple[str, str]]:
+    dup = set()
+    for label, rows in inv.items():
+        counts = Counter(r["op"] for r in rows)
+        dup |= {(label, op) for op, n in counts.items() if n > 1}
+    return dup
+
+
+def load_dismiss(path: Path) -> Set[Tuple[str, str]]:
+    if not path.is_file():
+        return set()
+    out = set()
+    for line in path.read_text().splitlines():
+        line = line.strip()
+        if not line or line.startswith("#"):
+            continue
+        repo, _, op = line.rpartition(":")
+        out.add((repo or "main", op))
+    return out
+
+
+def resolve_selection(args: argparse.Namespace,
+                      inv: Dict[str, List[Dict[str, str]]]) -> List[Case]:
+    dup = _dup_ops(inv)
+    index: Dict[Tuple[str, str, str], Dict[str, str]] = {}
+    by_op: Dict[Tuple[str, str], List[Dict[str, str]]] = {}
+    for label, rows in inv.items():
+        for r in rows:
+            index[(label, r["op"], r["op_file"])] = r
+            by_op.setdefault((label, r["op"]), []).append(r)
+
+    dismissed = load_dismiss(args.dismiss)
+
+    def make_case(row: Dict[str, str]) -> Case:
+        return Case(row["repo"], row["op"], row["op_file"], _run_name(row, dup))
+
+    if args.all:
+        cases = []
+        for label, rows in inv.items():
+            if args.repos and label not in args.repos:
+                continue
+            for r in rows:
+                if r.get("collected") == "no":
+                    continue
+                if r.get("marked_skip") == "yes" and not args.include_marked_skip:
+                    continue
+                if (label, r["op"]) in dismissed:
+                    continue
+                cases.append(make_case(r))
+        return cases
+
+    specs: List[str] = list(args.ops or [])
+    if args.ops_file:
+        for line in args.ops_file.read_text().splitlines():
+            line = line.strip()
+            if line and not line.startswith("#"):
+                specs.append(line)
+    if not specs:
+        raise SystemExit(f"{LOG} 未选择算子：用 --all / --ops / --ops-file 之一")
+
+    cases, errors = [], []
+    for spec in specs:
+        repo, _, rest = spec.rpartition(":")
+        repo = repo or "main"
+        op, _, op_file = rest.partition("@")
+        if repo not in inv:
+            errors.append(f"{spec}: 未知仓库 {repo!r}")
+            continue
+        if op_file:
+            row = index.get((repo, op, op_file))
+            if row is None:
+                errors.append(f"{spec}: inventory 中无此条目")
+                continue
+            cases.append(make_case(row))
+            continue
+        rows = by_op.get((repo, op)) or []
+        if not rows:
+            errors.append(f"{spec}: inventory 中无此算子")
+        elif len(rows) > 1:
+            files = ", ".join(r["op_file"] for r in rows)
+            errors.append(f"{spec}: 重名，需用 @op_file 消歧（候选: {files}）")
+        else:
+            cases.append(make_case(rows[0]))
+    if errors:
+        for e in errors:
+            print(f"{LOG} error: {e}", file=sys.stderr)
+        raise SystemExit(1)
+    return cases
+
+
+# ---------------- GPU 分片 ----------------
+
+def visible_gpus(arg: Optional[str]) -> List[str]:
+    if arg:
+        return [g.strip() for g in arg.split(",") if g.strip()]
+    env = os.environ.get("CUDA_VISIBLE_DEVICES", "").strip()
+    if env:
+        return [g.strip() for g in env.split(",") if g.strip()]
+    try:
+        out = subprocess.run(["nvidia-smi", "--list-gpus"], text=True,
+                             capture_output=True, timeout=30).stdout
+        n = sum(1 for line in out.splitlines() if line.startswith("GPU "))
+        if n:
+            return [str(i) for i in range(n)]
+    except Exception:
+        pass
+    return ["0"]
+
+
+def partition(cases: List[Case], gpus: List[str]) -> Dict[str, List[Case]]:
+    """按 (repo, run_name) 稳定哈希分片：与选择顺序无关，resume 时同卡复用。"""
+    parts: Dict[str, List[Case]] = {g: [] for g in gpus}
+    for c in cases:
+        digest = hashlib.sha256(f"{c.repo}/{c.run_name}".encode()).digest()
+        parts[gpus[int.from_bytes(digest[:8], "big") % len(gpus)]].append(c)
+    return parts
+
+
+# ---------------- 单算子执行 ----------------
+
+def _spawn(case: Case, out_dir: Path, gpu: str, args: argparse.Namespace,
+           dtypes: str) -> subprocess.Popen:
+    env = os.environ.copy()
+    env.pop("PARALLEL_WARMUP_GPUS", None)   # 批量层已分片，算子内不再并行
+    env.pop("REPLAY_FROM", None)
+    env.pop("OUT_DIR", None)
+    env.update({
+        "OP": case.op,
+        "OP_FILE": case.op_file,
+        "FLAGGEMS_DIR": REPO_DIRS[case.repo],
+        "OUT_DIR": str(out_dir),
+        "CUDA_VISIBLE_DEVICES": gpu,
+        "FLAGGEMS_PERF_COLOR": "never",
+        "NO_COLOR": "1",
+        "USE_FLAGTUNE": str(args.use_flagtune),
+    })
+    if args.shape_file:
+        env["SHAPE_FILE"] = str(args.shape_file)
+    if dtypes:
+        env["DTYPES"] = dtypes
+    if args.replay_root:
+        env["REPLAY_FROM"] = str(args.replay_root / case.repo / case.run_name)
+    return subprocess.Popen(
+        ["bash", str(ZL_BENCH / "run_pytest.sh")],
+        cwd=str(ZL_BENCH), env=env, start_new_session=True,
+        stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
+    )
+
+
+def _wait(proc: subprocess.Popen, timeout: int) -> Tuple[int, bool]:
+    """返回 (rc, timed_out)；超时对整个进程组 SIGTERM→SIGKILL。"""
+    try:
+        return proc.wait(timeout=timeout or None), False
+    except subprocess.TimeoutExpired:
+        try:
+            pgid = os.getpgid(proc.pid)
+            os.killpg(pgid, signal.SIGTERM)
+            try:
+                proc.wait(timeout=10)
+            except subprocess.TimeoutExpired:
+                os.killpg(pgid, signal.SIGKILL)
+        except ProcessLookupError:
+            pass  # 组内进程恰好在超时瞬间退出
+        proc.wait()
+        return 124, True
+
+
+def run_case(case: Case, batch_dir: Path, gpu: str,
+             args: argparse.Namespace) -> Dict[str, str]:
+    out_dir = batch_dir / case.repo / case.run_name
+    marker = out_dir / ".complete"
+
+    if args.resume and marker.is_file() and (out_dir / "run.log").is_file():
+        row = _classify(case, out_dir, gpu, 0, 0, "", timed_out=False)
+        row["notes"] = _join("resumed", row["notes"])
+        return row
+
+    out_dir.mkdir(parents=True, exist_ok=True)
+    marker.unlink(missing_ok=True)
+
+    dtypes = args.dtypes or ""
+    t0 = time.time()
+    rcs: List[int] = []
+    notes = ""
+    timed_out = False
+    for attempt in range(MAX_RETRIES + 1):
+        if attempt:
+            # 保留上一轮日志再重试
+            log = out_dir / "run.log"
+            if log.is_file():
+                log.rename(out_dir / f"run.attempt{attempt - 1}.log")
+        rc, timed_out = _wait(_spawn(case, out_dir, gpu, args, dtypes),
+                              args.op_timeout)
+        rcs.append(rc)
+        if timed_out:
+            break
+        log_text = _read_log(out_dir)
+        # dtype 不支持：去掉 --dtypes 降级重试（只降一次，不计入重试预算）
+        if rc != 0 and dtypes and DTYPE_UNSUPPORTED_MARK in log_text:
+            dtypes = ""
+            notes = _join(notes, "dtype_fallback")
+            continue
+        if rc == 0 or not (rc == 1 or rc >= 128):
+            break
+        if len(rcs) >= 2 and rcs[-1] == rcs[-2]:
+            break  # 确定性失败：同 rc 连续两次，不再烧预算
+
+    # dtype 营救：限制 dtype 时整体失败且没有任何成功行，去掉限制最后救一次。
+    # 覆盖 baseline 对受限 dtype 编译失败（torch jiterator 的 special 函数族等）
+    # 但报错文案不是上游 "can't be supported by this op" 的场景。
+    if dtypes and not timed_out:
+        probe = _classify(case, out_dir, gpu, rcs[-1], 0, "", False)
+        if probe["status"] == "FAIL" and probe["rows_success"] == "0":
+            log = out_dir / "run.log"
+            if log.is_file():
+                log.rename(out_dir / f"run.attempt{len(rcs) - 1}.log")
+            notes = _join(notes, "dtype_rescue")
+            rc, timed_out = _wait(_spawn(case, out_dir, gpu, args, ""),
+                                  args.op_timeout)
+            rcs.append(rc)
+    elapsed_ms = int((time.time() - t0) * 1000)
+
+    if len(rcs) > 1:
+        notes = _join(notes, "retry_rc=" + "->".join(map(str, rcs)))
+    row = _classify(case, out_dir, gpu, rcs[-1], elapsed_ms, notes, timed_out)
+    if row["status"] in ("PASS", "SKIP"):
+        tmp = marker.with_name(".complete.tmp")
+        tmp.write_text(row["status"] + "\n")
+        os.replace(tmp, marker)
+    return row
+
+
+def _read_log(out_dir: Path) -> str:
+    log = out_dir / "run.log"
+    return log.read_text(errors="replace") if log.is_file() else ""
+
+
+def _join(*notes: str) -> str:
+    return ";".join(n for n in notes if n)
+
+
+def _classify(case: Case, out_dir: Path, gpu: str, rc: int, elapsed_ms: int,
+              notes: str, timed_out: bool) -> Dict[str, str]:
+    text = _read_log(out_dir)
+    rows = batch_summary.parse_run_log(text)
+    if timed_out:
+        status = "TIMEOUT"
+    else:
+        status = batch_summary.classify(rc, rows, text)
+    n_skipped = batch_summary.pytest_skipped_count(text)
+    if status == "PASS" and n_skipped:
+        notes = _join(notes, f"partial_skip={n_skipped}")
+    return {
+        "repo": case.repo,
+        "op": case.op,
+        "op_file": case.op_file,
+        "run_name": case.run_name,
+        "gpu": gpu,
+        "status": status,
+        "return_code": str(rc),
+        "elapsed_ms": str(elapsed_ms),
+        "rows_total": str(len(rows)),
+        "rows_success": str(sum(1 for r in rows if r["row_status"] == "SUCCESS")),
+        "notes": _join(notes, batch_summary.marker_notes(text)),
+        "out_dir": str(out_dir),
+    }
+
+
+# ---------------- 主流程 ----------------
+
+def parse_args() -> argparse.Namespace:
+    p = argparse.ArgumentParser(
+        description=__doc__.splitlines()[0],
+        formatter_class=argparse.RawDescriptionHelpFormatter)
+    sel = p.add_argument_group("算子选择")
+    sel.add_argument("--all", action="store_true",
+                     help="跑 inventory 全量（排除 collected=no / marked_skip / dismiss）")
+    sel.add_argument("--ops", nargs="*", default=None, metavar="[repo:]op[@op_file]",
+                     help="直接指定算子（默认 main 仓；重名需 @op_file 消歧）")
+    sel.add_argument("--ops-file", type=Path, default=None,
+                     help="算子清单文件，每行一个 [repo:]op[@op_file]，# 注释")
+    sel.add_argument("--repos", nargs="*", default=None, choices=list(REPO_DIRS),
+                     help="--all 时限定仓库（默认全部）")
+    sel.add_argument("--dismiss", type=Path, default=OPS_DIR / "dismiss.txt",
+                     help="排除清单（默认 ops/dismiss.txt，仅 --all 生效）")
+    sel.add_argument("--include-marked-skip", action="store_true",
+                     help="--all 时不剔除带无条件 skip 标记的算子")
+
+    run = p.add_argument_group("执行")
+    run.add_argument("--gpus", default=None,
+                     help="逗号分隔 GPU 列表（默认 CUDA_VISIBLE_DEVICES 或全部）")
+    run.add_argument("--batch-dir", type=Path, default=None,
+                     help="产物目录（默认 runs/batch_<时间戳>；resume 时必须指定）")
+    run.add_argument("--op-timeout", type=int, default=1800,
+                     help="单算子超时秒数，0=不限（默认 %(default)s）")
+    run.add_argument("--resume", action="store_true",
+                     help="复用 --batch-dir 里带 .complete 标记的算子结果")
+    run.add_argument("--replay-root", type=Path, default=None,
+                     help="上次批量目录：逐算子 REPLAY_FROM <root>/<repo>/<op>")
+    run.add_argument("--shape-file", type=Path, default=OPS_DIR / "shapes_single.yaml",
+                     help="shape yaml（默认 %(default)s；深查用 ops/shapes_multi.yaml；"
+                          "传空串则用各算子上游默认 shape）")
+    run.add_argument("--dtypes", default="",
+                     help='限制 dtype 集（空格分隔，如 "bfloat16"）；空=上游默认扫描')
+    run.add_argument("--use-flagtune", type=int, default=0, choices=(0, 1),
+                     help="透传 USE_FLAGTUNE（默认 0；批量 A/B 两侧必须同值）")
+    args = p.parse_args()
+
+    if args.resume and not args.batch_dir:
+        raise SystemExit(f"{LOG} --resume 需要显式 --batch-dir")
+    if args.shape_file and str(args.shape_file).strip():
+        args.shape_file = args.shape_file.resolve()
+        if not args.shape_file.is_file():
+            raise SystemExit(f"{LOG} shape 文件不存在: {args.shape_file}")
+    else:
+        args.shape_file = None
+    if args.replay_root:
+        args.replay_root = args.replay_root.resolve()
+        if not args.replay_root.is_dir():
+            raise SystemExit(f"{LOG} replay 目录不存在: {args.replay_root}")
+    return args
+
+
+def main() -> None:
+    args = parse_args()
+    inv = load_inventory()
+    cases = resolve_selection(args, inv)
+    if not cases:
+        raise SystemExit(f"{LOG} 选择结果为空")
+
+    gpus = visible_gpus(args.gpus)
+    batch_dir = (args.batch_dir or
+                 ZL_BENCH / "runs" / f"batch_{datetime.now():%Y%m%d-%H%M%S}")
+    batch_dir = batch_dir.resolve()
+    batch_dir.mkdir(parents=True, exist_ok=True)
+
+    parts = {g: cs for g, cs in partition(cases, gpus).items() if cs}
+    print(f"{LOG} {len(cases)} 个算子 -> {len(parts)} 张卡 "
+          f"(shards: {[len(cs) for cs in parts.values()]})  out={batch_dir}")
+    if args.replay_root:
+        print(f"{LOG} replay 模式: REPLAY_FROM={args.replay_root}/<repo>/<op>")
+
+    lock = threading.Lock()
+    done = [0]
+    results: Dict[str, Dict[str, str]] = {}
+    status_csv = batch_dir / "ops_status.csv"
+    wname = max((len(c.run_name) for c in cases), default=8)
+
+    def flush_status() -> None:
+        rows = [results[c.key] for c in cases if c.key in results]
+        tmp = status_csv.with_name(status_csv.name + ".tmp")
+        with tmp.open("w", newline="") as f:
+            writer = csv.DictWriter(f, fieldnames=STATUS_FIELDS)
+            writer.writeheader()
+            writer.writerows(rows)
+        os.replace(tmp, status_csv)
+
+    def run_partition(gpu: str, part: List[Case]) -> None:
+        for case in part:
+            try:
+                row = run_case(case, batch_dir, gpu, args)
+            except Exception as exc:  # 编排层异常不拖垮整卡分片
+                row = {
+                    "repo": case.repo, "op": case.op, "op_file": case.op_file,
+                    "run_name": case.run_name, "gpu": gpu, "status": "FAIL",
+                    "return_code": "EXC", "elapsed_ms": "0", "rows_total": "0",
+                    "rows_success": "0", "notes": f"harness_exc={exc}",
+                    "out_dir": str(batch_dir / case.repo / case.run_name),
+                }
+            with lock:
+                done[0] += 1
+                results[case.key] = row
+                flush_status()
+                print(f"  ({done[0]:>4}/{len(cases)})  {row['status']:<7} "
+                      f"{row['elapsed_ms']:>8}ms  rows={row['rows_success']}/"
+                      f"{row['rows_total']:<4} {case.repo}:{case.run_name:<{wname}} "
+                      f"[gpu{gpu}]"
+                      + (f"  {row['notes']}" if row["notes"] else ""),
+                      flush=True)
+
+    if len(parts) == 1:
+        gpu, part = next(iter(parts.items()))
+        run_partition(gpu, part)
+    else:
+        with ThreadPoolExecutor(max_workers=len(parts)) as pool:
+            futures = [pool.submit(run_partition, g, p) for g, p in parts.items()]
+            for f in futures:
+                f.result()
+
+    summary = batch_summary.summarize(batch_dir)
+    rows = [results[c.key] for c in cases if c.key in results]
+    counts = {s: sum(1 for r in rows if r["status"] == s)
+              for s in ("PASS", "SKIP", "FAIL", "TIMEOUT")}
+    print(f"{LOG} 完成: " + "  ".join(f"{k}={v}" for k, v in counts.items()))
+    print(f"{LOG} 状态表: {status_csv}")
+    print(f"{LOG} 总表:   {summary}")
+    bad = [r for r in rows if r["status"] in ("FAIL", "TIMEOUT")]
+    if bad:
+        print(f"{LOG} 失败算子: " +
+              ", ".join(f"{r['repo']}:{r['run_name']}" for r in bad))
+        sys.exit(1)
+
+
+if __name__ == "__main__":
+    main()
diff --git a/run_pytest.sh b/run_pytest.sh
index c5a04e6..f6a5540 100644
--- a/run_pytest.sh
+++ b/run_pytest.sh
@@ -42,6 +42,10 @@ YAML
 # 调优空间：0=普通 autotune（默认，快速验证）；1=FlagTune 扩展空间（首跑全量搜索、慢）
 USE_FLAGTUNE="${USE_FLAGTUNE:-0}"
 
+# 可选：限制 dtype 集（空格分隔，如 "bfloat16 float16"）。空=上游默认 dtype 扫描。
+# 算子不支持指定 dtype 时上游会报 "can't be supported by this op"（批量驱动据此降级重试）。
+DTYPES="${DTYPES:-}"
+
 # 空=record 模式，把本次选中的 config 记入 autotune_records/<op>.json；
 # 指向某次历史 run 目录则 replay 其记录，用于 A/B 两侧锁同一套 config。
 REPLAY_FROM="${REPLAY_FROM:-}"
@@ -104,6 +108,12 @@ export PYTHONPATH="$SCRIPT_DIR${PYTHONPATH:+:$PYTHONPATH}"
 export FLAGGEMS_PERF_CURRENT_OP="$OP"
 export PYTHONUNBUFFERED=1  # 实时输出不缓冲
 
+# DTYPES 非空时逐个转为上游 --dtypes 选项（action=append，每个 dtype 一次）
+DTYPE_ARGS=()
+for _dt in $DTYPES; do
+  DTYPE_ARGS+=(--dtypes "$_dt")
+done
+
 # record/replay 互斥，各由自己的环境变量激活
 if [[ -n "$REPLAY_FROM" ]]; then
   AUTOTUNE_ENV="FLAGGEMS_PERF_AUTOTUNE_REPLAY_DIR=$REPLAY_FROM/autotune_records"
@@ -139,7 +149,7 @@ status=0
   env "$AUTOTUNE_ENV" \
   USE_FLAGTUNE=$USE_FLAGTUNE python -u -m pytest -s "$TEST_FILE" \
     "${PLUGINS[@]}" "${PYTEST_COLOR[@]}" \
-    --shape_file "$SHAPE_FILE" \
+    --shape_file "$SHAPE_FILE" ${DTYPE_ARGS[@]+"${DTYPE_ARGS[@]}"} \
     --level core --mode kernel || status=$?
 
   rm -rf "$CACHE_DIR"