Commit Graph

  • c0eba77ed6 switch backend to setuptools main Jinjie Liu 2026-04-20 10:23:36 +08:00
  • a727526794 add length check to unwrap arguments Jinjie Liu 2026-02-12 15:31:50 +08:00
  • 599957e156 support attention bwd Jinjie Liu 2026-02-10 17:01:28 +08:00
  • e41ec26329 support attention fwd Jinjie Liu 2026-02-09 17:15:51 +08:00
  • 213e4fc060 use templates to substitute parts of macros jinjieliu 2026-02-08 22:24:12 +08:00
  • 1c4f13c8f0 find packages by sysconfig instead of importlib jinjieliu 2026-02-08 16:15:56 +08:00
  • 24237a6313 include header files by c/cpp instead of jinja jinjieliu 2026-02-07 17:16:49 +08:00
  • 6a19a6b06d put num_warps and num_stages in kwargs jinjieliu 2026-02-07 14:25:10 +08:00
  • 2298b6f8c8 support mm and autotune jinjieliu 2026-02-07 00:41:23 +08:00
  • f6c7a48c1b enable lambda function for grid descriptor Jinjie Liu 2026-02-05 15:59:22 +08:00
  • 8b8aa6cb84 enable optional for numwarps and numstages jinjieliu 2026-02-05 01:01:49 +08:00
  • b7bf598fde support softmax Jinjie Liu 2026-02-04 16:38:33 +08:00
  • 192dc95ac0 supports decorator for jit and wrapper Jinjie Liu 2026-02-04 10:46:14 +08:00
  • 6e4c2d4a43 fix bugs on lacking cudatoolkit Jinjie Liu 2026-02-04 10:27:44 +08:00
  • dc8c2c17e0 verify tvm-ffi cpp wrapper on vector-add.py jinjieliu 2026-02-04 02:30:26 +08:00