pytorch/benchmark

Pull Requests Commits

FindHao committed 1 year ago

8c9689f1

add multiple ops support for --op argument

FindHao committed 1 year ago

05067160

Enable fp8 rowwise on AMDGPU (#2483)

htyu committed 1 year ago

f2932b74

Add layout options to gemm

bertmaher committed 1 year ago

4445aa2b

CSV of extra shapes for gemm benchmarks

bertmaher committed 1 year ago

d512e673

Modernize cutlass call for fp8 blockwise

bertmaher committed 1 year ago

2d9ab0b1

Print TMA benchmark info to stderr

bertmaher committed 1 year ago

0cb1e96d

Skip loading triton.nvidia.cublas if not found

bertmaher committed 1 year ago

0611c41c

Make FA3 work in fbcode

bertmaher committed 1 year ago

b6b67a4f

Fix hardcoded shape in low_mem_dropout benchmark (#2475)

mark14wu committed 1 year ago

252a3b17

Restore FlexAttention and FlashV3 backward (#2473)

bertmaher committed 1 year ago

611bf702

Fix bug #2458 (#2459)

ostrowskimarcin committed 1 year ago

0f050151

Add nsys integration

xuzhao9 committed 1 year ago

2edf80cb

Revert "Trace enter/exit of TorchFunctionModes (#135422)" (#136590)

anijain2305 committed 1 year ago

a31c3fe9

Move FBGEM to 5c3d54f335b1617b5b169061c3b3d59b2a791ebb (#2471)

htyu committed 1 year ago

8a690690

Do not treat user defined nn module attributes static for dynamic shape infra (#136516)

anijain2305 committed 1 year ago

93b58047

Add release tests (#2470)

atalman committed 1 year ago

7d30648b

Add types to _dynamo/testing.py (#136402)

bobrenjc93 committed 1 year ago

deb2988e

Add HSTU ragged attention operator (#2453)

xuzhao9 committed 1 year ago

926fb10a

bf16xint16_gemm operator: add --transpose option (#2466)

davidberard98 committed 1 year ago

0ab0e47e

Move FBGEM to fcf36b445a1929c8a22f07c35d3a29b3697867a4 (#2467)

htyu committed 1 year ago

6a089a45

Add _dynamo.config.suppress_errors logging (#136379)

jovianjaison committed 1 year ago

8d57e930

Run single iteration when collecting ncu traces

xuzhao9 committed 1 year ago

f2f0b30f

make --dump-ir also dump sass (#2460)

davidberard98 committed 1 year ago

46ab2e2b

Log structured logging overhead to dynamo compile (kinda) (#2454)

jamesjwu committed 1 year ago

9c7aacc4

Increase multiplier to 3 for Inductor AMP FP16 benchmark correctness check (#135932)

jiayisunx committed 1 year ago

c1755f53

use npt.NDArray instead of np.ndarray in type annotations

igorsugak committed 1 year ago

c97859eb

Update mypy to 1.11.2 (#133816)

Skylion007 committed 1 year ago

7e1ba8d5

Remove ignored modes workaround (#135502)

mlazos committed 1 year ago

6b1c40bc

Trace enter/exit of TorchFunctionModes (#135422)

mlazos committed 1 year ago

99ea924e

Older