strisunshinewentingwang/pytorch

Pull Requests Commits

[Inductor] Add an option to mark wrapper call in PyTorch profiler (#89674)

jgong5 committed 1 year ago

c75434ed

[functorch] fix possible overflow (#83389)

cyyever committed 1 year ago

4b11119c

Fix archive issue impacting summary stat diff (#89789)

williamwen42 committed 1 year ago

63843401

[FSDP] Fix `nn.Parameter` usage for 2D and `use_orig_params=True` (#89782)

awgu committed 1 year ago

943acd4d

[Checkpoint][2D][4/N] Add nested_dict for distributed checkpoint to core distributed (#89537)

wz337 committed 1 year ago

23ee6757

Re-enabled 3 reductions tests on Windows (#89567)

mantaionut committed 1 year ago

a378ba21

Add bits tensor types (#88594)

angelayi committed 1 year ago

f3b1315e

[Checkpoint][2D][3/N] Add nested_tensors for distributed checkpoint to core distributed (#89501)

wz337 committed 1 year ago

22e7514a

[CUDA graphs] Add warning if captured graph is empty (#88754)

Aidyn-A committed 1 year ago

0057be33

[skip ci] documentation update for the kwargs defaults section of fun… (#89719)

weberxie committed 1 year ago

c18da597

[LTC] Metrics can be reset too (#89606)

alanwaketan committed 1 year ago

13d2af2a

[Vulkan][TCC] Fix conv2d pack biases (#89568)

manuelcandales committed 1 year ago

5abe454d

Add meta implementation for _efficientzerotensor (#88936)

peterbell10 committed 1 year ago

2e0cd7c8

Fix comparison of batched_prop vs unbatched_prob in test_distributions (#87977)

Flamefire committed 1 year ago

69a8c92d

Revert "Move Dynamo docs back to core (#89769)"

pytorchmergebot committed 1 year ago

47cca5e4

Tweak formatting of note on macros (#89598)

eqy committed 1 year ago

83210660

Move Dynamo docs back to core (#89769)

msaroufim committed 1 year ago

be2816db

[inductor] skip dm_nfnet_f0 in TIMM model test (#89768)

desertfire committed 1 year ago

465ee7bc

[benchmarks] Disabling gradscaler (#89741)

anijain2305 committed 1 year ago

cdf40875

Revert "Don't allow recomputing a node that *must* be materialized in the backwards pass (#89171)" (#89770)

ezyang committed 1 year ago

e8643ded

[multipy] Address GetPythonFramesFunction() and multipy incompatibility. (#267) (#89315)

PaliC committed 1 year ago

2a2c07ae

Reland "Add single process version of dynamo distributed hf_Bert tests (#89721)" (#89756)

ezyang committed 1 year ago

95563b3e

Revert "Support set_rng_state with fake tensor (#89642)"

pytorchmergebot committed 1 year ago

6ef70249

Update minor release acceptance criteria (#89767)

atalman committed 1 year ago

ed41a7fb

Add AOTAutograd and partitioner to ciflow/inductor (#89772)

ezyang committed 1 year ago

ed9cd47e

Use isinstance test rather than exact type test for wrap to fake (#89671)

ezyang committed 1 year ago

cf91e364

Make aot_module_simplified accept fake tensors (#89670)

ezyang committed 1 year ago

b87c45d5

Change aot_module_simplified to take take arguments directly (#89669)

ezyang committed 1 year ago

abf91562

Refactor how AOTAutograd backends are defined (#89736)

ezyang committed 1 year ago

b589e726

[ROCm] Replace layer_norm_grad_input_kernel with cuComputeGradInput for ROCm (#87726)

hubertlu-tw committed 1 year ago

cf4969d9

Newer Older