pytorch/benchmark

Pull Requests Commits

Enable emulate_precision_casts for mobilenetv2_100. (#187364)

generatedunixname499836121 committed 1 day ago

c8bca312

Fix Dynamo set operations over pre-existing generators (#186042)

generatedunixname499836121 committed 7 days ago

a1c3f0e7

Fix timm fp16 accuracy for new models (#185038)

jansel committed 8 days ago

2dbbb097

Fix timm benchmark dependency installation (#186289)

jansel committed 8 days ago

b0b38eea

Fix Dynamo sequence membership to use richcompare per element (#186760)

generatedunixname499836121 committed 12 days ago

811f78de

Enable `emulate_precision_casts` for `vit_base_patch14_dinov2.lvd142m` (#185676)

generatedunixname499836121 committed 13 days ago

008bc712

Fix bool mask IoU comparison for Detectron2 outputs (#186441)

jansel committed 14 days ago

20e03326

Disable fallback_random for DistillGPT2 accuracy/training (#186623)

generatedunixname499836121 committed 15 days ago

83514f6f

disable nested graph breaks on inline_user_function_return by default (#179844)

williamwen42 committed 19 days ago

4bb7da61

tp_richcompare followups (#184606)

williamwen42 committed 20 days ago

5786e143

Add TensorVariable/RangeVariable to unpack_iterable fast path; remove tensor branch from UNPACK_SEQUENCE (#185152)

generatedunixname499836121 committed 22 days ago

3c87e5be

Avoid Triton 256MB autotune cache allocation (#184479)

jansel committed 22 days ago

d0d0d40c

Trace through einops 0.8.2 in Dynamo (#185619)

jansel committed 22 days ago

c708a21d

Fix scoped Chromium event reset (#184973)

jansel committed 24 days ago

126c79d4

Improve torch.compile device mismatch errors (#185412)

jansel committed 24 days ago

bd5b67d9

change InstructionTranslator typing to InstructionTranslatorBase (#183577)

williamwen42 committed 26 days ago

2b5cfc1e

Make dynamo debug/repro utilities device-agnostic (#184851)

generatedunixname499836121 committed 26 days ago

f65725fa

refactor(dynamo): Implement tp_repr via generic_repr (#182332)

generatedunixname499836121 committed 27 days ago

6c28f08d

mimic tp_richcompare handling (#182759)

williamwen42 committed 28 days ago

171d7914

Revert "mimic tp_richcompare handling (#182759)"

generatedunixname499836121 committed 29 days ago

4ddce46a

Raise IndexError compile mode as Eager mode - Again after Revert (#184856)

generatedunixname499836121 committed 31 days ago

e18df6e7

mimic tp_richcompare handling (#182759)

williamwen42 committed 31 days ago

cdc4ae03

Support AOTAutograd cache keys for NumPy wrappers (#184074)

jansel committed 34 days ago

a7177268

Refactor BuiltinVariable methods to use getiter / getiter_next for iterable handling (#181116)

generatedunixname499836121 committed 34 days ago

3568dd68

reland pipe dynamic spec in dynamo torch.compile (#2691)

laithsakka committed 35 days ago

07a42098

Improve Dynamo output graph metadata typing (#184369)

bobrenjc93 committed 35 days ago

19d7935b

Skip Dynamo graph break for scalar-only bin_ops when tensorify is enabled

nandesuka committed 36 days ago

4713e3b9

Revert D105603961: pipe dynamic spec in dynamo torch.compile (#181995)

izaitsevfb committed 37 days ago

ec4b4866

pipe dynamic spec in dynamo torch.compile (#181995)

laithsakka committed 37 days ago

de691a8c

Log functorch_config to Scuba per compile region (#182762) (#182762)

jackietrang committed 41 days ago

a3ca6fca

Older