pytorch/benchmark

Pull Requests Commits

Add tests for B200

huydhn committed 1 year ago

7d75608b

Restore TorchBench PR tests (#2636)

huydhn committed 1 year ago

9ebf2e28

Add torchscript jit.trace to benchmark option (#161223)

yiming0416 committed 1 year ago

9644652f

Log exception_stack_trace to dynamo_compile (#161096)

jovianjaison committed 1 year ago

78745baf

Revamp TorchBench nightly image build process (#2635)

huydhn committed 1 year ago

b0efcbe2

Refactor convert_frame._compile_inner to return compiled bytecode + output graph. [3/n] (#160855)

zhxchen17 committed 1 year ago

abd8dd23

Add caching for inside torch.compile.disable function to avoid unnecessary recompilation. (#160934)

generatedunixname499836121 committed 1 year ago

3e377c41

Type coverage for symbolic_convert (#160922)

Lucaskabela committed 1 year ago

73eb12ed

Cleanup unused TorchBench workflows (#2634)

huydhn committed 1 year ago

9a4b981e

Revert "Fix the benchmark CI failure due to Opacus 1.5.4 (#2621)" (#2632)

huydhn committed 1 year ago

22bc29b4

pack linear for FP32 dynamic mode (#157542)

generatedunixname499836121 committed 1 year ago

af93b8be

Add GPT2ForSequenceClassification to use_larger_multiplier_for_smaller_tensor list (#160001)

generatedunixname499836121 committed 1 year ago

9e418359

Add logging for when inbuilt_inline_nn_modules will help with ID_MATCH guard triggered recompiles (#160592)

ppanchalia committed 1 year ago

6da75d11

Type improvements in `_dynamo/utils` to generics (#159824)

Lucaskabela committed 1 year ago

24586097

Log graph_node_shape to dynamo_compile (#160556)

jovianjaison committed 1 year ago

e11cbaa1

Make yolov3 exportable (#2630)

tugsbayasgalan committed 1 year ago

290c966f

Replace `pytorch-labs` with `meta-pytorch` (#2629)

ZainRizvi committed 1 year ago

136d13a9

Log stack_trace to dynamo_compile (#160348)

jovianjaison committed 1 year ago

a9216941

Bump transformers pin (#159291)

anijain2305 committed 1 year ago

13ac4f7c

Add nativert benchmark (#159922)

yiming0416 committed 1 year ago

c9ccd888

Typing `torch/_dynamo/guards.py` (#159315)

Lucaskabela committed 1 year ago

ac742a2b

Type annotation for `torch/_dynamo/output_graph.py` (#159602)

Lucaskabela committed 1 year ago

5bbf4e8c

Type coverage for `torch/_dynamo/utils.py` (#159580)

Lucaskabela committed 1 year ago

903b5003

gc before warming up benchmarking (#159670)

aorenste committed 1 year ago

3c500130

Enable sample nightly PT2 benchmark on B200 (#158011)

huydhn committed 1 year ago

b4dd0684

Add typing annotations to guard and source

Lucaskabela committed 1 year ago

fdea2a14

Support native Inductor as backend for MTIA

andyanwang committed 1 year ago

28b27b9e

Set model name early to keep warmup and main model same (#159231)

anijain2305 committed 1 year ago

b377227f

Revert D78458745: Support native Inductor as backend for MTIA

ZainRizvi committed 1 year ago

bed0bfe6

Support native Inductor as backend for MTIA

andyanwang committed 1 year ago

6a1b132d

Older