pytorch/benchmark

Pull Requests Commits

Remove hammer/generative_recommenders (#2526)

xing-liu committed 1 year ago

4d6e0fa0

Set use_cuda_graphs in fp8_gemm_rowwise

karthik-man committed 1 year ago

870be9b4

add some cpython debugging methods (#138030)

williamwen42 committed 1 year ago

438f82b4

Switch times to us in CompilationMetrics and improvements (#138975)

ezyang committed 1 year ago

4ad2712d

Back out "tls access helpers (#138061)" and Back out "[compiled autograd] Compiled autograd configs in TLS (#137821)"

xmfan committed 1 year ago

47e3138d

Fix naming for AMD in fp8 rowwise fbgemm

karthik-man committed 1 year ago

713f8002

Use guard_manager consistently instead of check_fn (#138896)

anijain2305 committed 1 year ago

34ea1a1c

Update Typeguard to TypeIs for better type inference (#133814)

Skylion007 committed 1 year ago

bd238116

pytorch/benchmark:utils

generatedunixname89002005287564 committed 1 year ago

f8a4e518

pytorch/benchmark:bisection

generatedunixname89002005287564 committed 1 year ago

f6f1249c

Add logger logging for remote fx graph cache get + put (#2512)

masnesral committed 1 year ago

2614ca98

fixing gemm for amd

adamomainz committed 1 year ago

65e5f686

Account for older numpy versions in #2514 (#2524)

mikaylagawarecki committed 1 year ago

4f30c497

Release 2.5.1.yaml perf test (#2525)

atalman committed 1 year ago

47ba1ed8

updating hardware and device columns

adamomainz committed 1 year ago

3245fde9

Support overridden __call__ on nn modules (#138619)

anijain2305 committed 1 year ago

b5b342ba

Support range_iterator as a function input (#138657)

anijain2305 committed 1 year ago

367b6ef3

adding fp32 strict and tf32x3 benchmarks for gemm

adamomainz committed 1 year ago

036012ff

tls access helpers (#138061)

xmfan committed 1 year ago

405ba75b

Compiled autograd configs in TLS (#137821)

xmfan committed 1 year ago

0e038319

Log is_forward field to dynamo_compile scuba table (#138505)

masnesral committed 1 year ago

e57bbe23

Add metadata to events in progress, new `dynamo` event

jamesjwu committed 1 year ago

8fce9c12

typing compile_fx.py (#138033)

aorenste committed 1 year ago

11543183

Fixes to prep for weights_only default flip (#2514)

mikaylagawarecki committed 1 year ago

fb590d99

Update requirements.txt (#2523)

wdvr committed 1 year ago

a45e0dbf

Replace __str__ with __repr__ in some places (#136316)

rec committed 1 year ago

173774d1

fixing key error in aggregate data

adamomainz committed 1 year ago

a21b30e4

Disable torch function compilation during guard execution and in compiled bytecode (#137669)

mlazos committed 1 year ago

05620407

replace uses of np.ndarray with npt.NDArray

igorsugak committed 1 year ago

06e35fc5

specifying logged benchmark name for tritonBench servicelab logging

adamomainz committed 1 year ago

e737b8fe

Newer Older