Commits

Branch selector

 master

User selector

yanboliang

Datepicker

All time

Commit History

Commits on Aug 16, 2024

• [Dynamo] Support torch.cuda.device ctx manager (pytorch#133385)

  Show description for 770086f
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  770086f

  Copy full SHA for 770086f

Commits on Aug 14, 2024

• [Inductor][FlexAttention] Support non-divisible sequence lengths (pytorch#133019)

  Show description for 5274104
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  5274104

  Copy full SHA for 5274104

Commits on Aug 13, 2024

• [Dynamo] Make torch.Size can be reconstructed by LOAD_CONST (pytorch#133342)

  Show description for 9de023d
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  9de023d

  Copy full SHA for 9de023d

Commits on Aug 3, 2024

• [Inductor][FlexAttention] TestFlexAttention -> TestFlexDecoding (pytorch#132547)

  Show description for bbce517
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  bbce517

  Copy full SHA for bbce517

• [Inductor][FlexAttention] Add kwarg to top level for users to specify kernel params (pytorch#132015)

  Show description for 373e9be
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  373e9be

  Copy full SHA for 373e9be

Commits on Aug 2, 2024

• [Dynamo] Support abc.MutableMapping.get (pytorch#132363)

  Show description for 5ea0f51
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  5ea0f51

  Copy full SHA for 5ea0f51

Commits on Jul 30, 2024

• [Inductor][FlexAttention] Correct partial/full blocks naming (pytorch#131993)

  Show description for 54d4f6b
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  54d4f6b

  Copy full SHA for 54d4f6b

Commits on Jul 29, 2024

• [BE] Fix torch.compile docstring formatting issues (pytorch#131837)

  Show description for 3c15621
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  3c15621

  Copy full SHA for 3c15621

Commits on Jul 26, 2024

• [Dynamo] Support zip_longest (pytorch#131497)

  Show description for e76e566
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  e76e566

  Copy full SHA for e76e566

• [Dynamo] Fix guard_on_nn_modules unit tests discrepancy between OSS and fbcode (pytorch#131810)

  Show description for b893a57
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  b893a57

  Copy full SHA for b893a57

Commits on Jul 25, 2024

• [Inductor] Added and_masks and or_masks utilities & make fully masked out rows 0 instead of nan (pytorch#131552)

  Show description for a34692c
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  a34692c

  Copy full SHA for a34692c

• [BE] Annotate subgraph_lowering (pytorch#131545)

  Show description for 5f3f14e
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  5f3f14e

  Copy full SHA for 5f3f14e

Commits on Jul 12, 2024

• [Inductor] FlexAttention supports partial masking (pytorch#130415)

  Show description for 207564b
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  207564b

  Copy full SHA for 207564b

Commits on Jul 4, 2024

• [Dynamo] Add assertion for tensor unpack shape mismatch (pytorch#130077)

  Show description for 551f3b9
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  551f3b9

  Copy full SHA for 551f3b9

Commits on Jul 3, 2024

• [Dynamo] Add config to skip/inline torchrec (pytorch#129912)

  Show description for 111f9b5
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  111f9b5

  Copy full SHA for 111f9b5

Commits on Jul 2, 2024

• [Inductor][FlexAttention] Add helper functions of converting score_mod to block_mask (pytorch#129909)

  Show description for 080149c
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  080149c

  Copy full SHA for 080149c

• [Inductor] Rename TemplatedAttention to FlexAttention (pytorch#129859)

  Show description for 1f3e2d7
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  1f3e2d7

  Copy full SHA for 1f3e2d7

• [Inductor] Make FlexAttention block_mask argument as tuple (pytorch#129831)

  Show description for 34e94c5
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  34e94c5

  Copy full SHA for 34e94c5

Commits on Jun 29, 2024

• [Inductor] FlexAttention supports block sparse mask (pytorch#129216)

  Show description for ec47d4d
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  ec47d4d

  Copy full SHA for ec47d4d

• [GPT-fast] Update micro benchmark numbers as A100-50G (pytorch#129799)

  Show description for 7b5a842
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  7b5a842

  Copy full SHA for 7b5a842

Commits on Jun 28, 2024

• [Inductor] FlexAttention supports block sparse mask (pytorch#129216)

  Show description for b9d3ced
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  b9d3ced

  Copy full SHA for b9d3ced

Commits on Jun 27, 2024

• [Inductor] FlexAttention supports block sparse mask (pytorch#129216)

  Show description for 4082759
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  4082759

  Copy full SHA for 4082759

Commits on Jun 26, 2024

• [GPT-benchmark] Distinguish LLM models and mirco-benchmarks (pytorch#129498)

  Show description for 9554a9a
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  9554a9a

  Copy full SHA for 9554a9a

Commits on Jun 19, 2024

• [torch.compile] Enable bwd compilation metrics (pytorch#128973)

  Show description for acefc5c
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  acefc5c

  Copy full SHA for acefc5c

Commits on Jun 17, 2024

• [Inductor][FlexAttention] Tune backwards kernel block sizes (pytorch#128853)

  Show description for 8953725
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  8953725

  Copy full SHA for 8953725

• [GPT-benchmark] Fix memory bandwidth for MoE (pytorch#128783)

  Show description for a489792
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  a489792

  Copy full SHA for a489792

• [GPT-benchmark] Add metric: compilation time for GPT models (pytorch#128768)

  Show description for 8c06eae
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  8c06eae

  Copy full SHA for 8c06eae

Commits on Jun 14, 2024

• [GPT-fast benchmark] Add MLP, gather + gemv, gemv micro benchmark (pytorch#128002)

  Show description for 1fb4eff
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  1fb4eff

  Copy full SHA for 1fb4eff

Commits on Jun 4, 2024

• [Inductor] Add FlexAttention backward kernel dynamic shape tests (pytorch#127728)

  Show description for 8d153e0
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  8d153e0

  Copy full SHA for 8d153e0

• [Inductor][Flex-attention] Support different sequence lengths for Query and Key/Value (pytorch#127678)

  Show description for e793ae2
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  e793ae2

  Copy full SHA for e793ae2

• [Inductor] FlexAttention backward kernel optimization (pytorch#127208)

  Show description for 69f5b66
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  69f5b66

  Copy full SHA for 69f5b66

Commits on Jun 3, 2024

• [Inductor] Add FlexAttention backward kernel dynamic shape tests (pytorch#127728)

  Show description for 10e3406
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  10e3406

  Copy full SHA for 10e3406

• [Inductor][Flex-attention] Support different sequence lengths for Query and Key/Value (pytorch#127678)

  Show description for e2e3ca9
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  e2e3ca9

  Copy full SHA for e2e3ca9

• [Dynamo] Log backward graph compilation metrics (pytorch#126629)

  Show description for 7e97b33
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  7e97b33

  Copy full SHA for 7e97b33

Commits on May 31, 2024

• [Inductor] FlexAttention backward kernel optimization (pytorch#127208)

  Show description for f717131
  

  yanboliang

  authored and

  pytorchmergebot

  committed
  f717131

  Copy full SHA for f717131

Pagination

Previous

Next