strisunshinewentingwang/pytorch

Pull Requests Commits

Don't unsafely clone autograd meta (#89720)

ezyang committed 1 year ago

6c1fb3f2

Fix CopySlices logic to ensure wrapped node runs properly. (#89812)

albanD committed 1 year ago

02e2eaa9

[test_nn] split multihead_attention from test_nn (#89748)

kshitij12345 committed 1 year ago

8314d403

[Quant][docs] Use get_default_qconfig_mapping (#87299)

andrewor14 committed 1 year ago

fb47a669

[Quant][fx][bc-breaking] Remove backend_config_utils.py (#89810)

andrewor14 committed 1 year ago

2bce6d09

Revert "[GHA] Decrease Windows test timeout to 120 minutes (#89694)"

pytorchmergebot committed 1 year ago

e1dbd9a2

[FSDP] Relax post-backward assert (#89791)

awgu committed 1 year ago

6e2da426

Revert "Move functorch/_src to torch/_functorch (#88756)"

pytorchmergebot committed 1 year ago

218d9c6e

[follow-up] Python Attr Serialization (#88913)

kshitij12345 committed 1 year ago

086b251f

don't run input mutation analysis in dynamo (#89760)

bdhirsh committed 1 year ago

2f9ec226

[aarch64] add SLEEF dependency for aten_cpu (#89475)

ajaymh committed 1 year ago

3cef87f9

[Quant][docs] Fix BackendConfig example in docstring/README (#89319)

andrewor14 committed 1 year ago

c6ede0bd

Move functorch/_src to torch/_functorch (#88756)

zou3519 committed 1 year ago

52bc5c1c

Guard the boundary of index computed in compute_source_index_and_lambda (#89252)

jgong5 committed 1 year ago

620994cd

[PyTorch Edge] Set training for module only (#89488)

cccclai committed 1 year ago

93772305

Refactoring to share vectorization code for int8/uint8. (#89650)

alexsamardzic committed 1 year ago

a78467f3

[minifier] Continue on assertion for accuracy minification (#89739)

anijain2305 committed 1 year ago

8226a5d3

[dynamo] Don't copy the graph during checkpointing (copy_graphstate) (#89232)

mlazos committed 1 year ago

40dd03ee

add memory_tracker tool to help profiling memory usages (#88825)

zhaojuanmao committed 1 year ago

91899a9e

Test FSDP with submodule non-reentrant checkpointing (#89781)

mrshenli committed 1 year ago

7ec7a820

Dynamo asserts FSDP wrapped modules use_orig_param (#89523)

wconstab committed 1 year ago

705ad36c

Enable DDPOptimizer by default in dynamo (#88523)

wconstab committed 1 year ago

7860fcc2

Move Dynamo docs back to core (#89769)

msaroufim committed 1 year ago

9048cf16

[dynamo] Minifier fixes for reproducing segfault (#89712)

anijain2305 committed 1 year ago

2b522670

[decomp] Fix native_batch_norm_backward dtype of dweight and dbias (#89740)

anijain2305 committed 1 year ago

c1950620

Call _sdp_attention in nn.functional.mha (#89470)

drisspg committed 1 year ago

4d7ec302

fixes for inductor <> batch norm (#89603)

bdhirsh committed 1 year ago

e20ec445

Add type hint to torch.norm and Tensor.norm (#89728)

fd0r committed 1 year ago

740860d4

[nvfuser] avoid out of bounds error (#89584)

davidberard98 committed 1 year ago

908daa8a

Special-case fsdp wrapped modules to be Unspecialized (#89330)

wconstab committed 1 year ago

77df2ca9

Newer Older