8000 Commits · tombousso/pytorch · GitHub

More Web Proxy on the site http://driver.im/

Commits

Commits on Jun 8, 2024

[dynamo] Skip inlining builtin nn modules for torch.compile inside cond (#128247)

anijain2305
authored and
pytorchmergebot
committed
[easy][inline-inbuilt-nn-modules] Fix expected graph for control flow test (#128246)

anijain2305
authored and
pytorchmergebot
committed

Commits on Jun 7, 2024

[inductor][dynamo-inline-nn-modules] Fix test with inlining flag (#128200)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Inline the getattr of fx graph and proxy graph (#128172)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Support if cond on UnspecializedNNModuleVariable and add inline tests (#128158)

anijain2305
authored and
pytorchmergebot
committed

Commits on Jun 6, 2024

[dynamo][nn-modules] Trace through nn.Module dunder methods for UnspecializedNNModule (#126578)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Bugfix for nn parameter construction (#128001)

anijain2305
authored and
pytorchmergebot
committed

Commits on Jun 5, 2024

[fake tensor] Set _is_param for base fake tensors for views (#127823)

anijain2305
authored and
pytorchmergebot
committed
[dynamo][pippy] Hotfix for nn_module_stack for pippy usecase (#127972)

anijain2305
authored and
pytorchmergebot
committed

Commits on Jun 4, 2024

[dynamo] Bugfix for nn parameter construction (#127806)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Unspec nn module when global backward hooks are present (#127802)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Tensorvariable - track grad with _grad field (#127785)

anijain2305
authored and
pytorchmergebot
committed

Commits on Jun 1, 2024

[dynamo] Support __getitem__ on NNModuleVariable __dict__ (#126956)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Initial exception handling support (#126923)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 31, 2024

[dynamo] Add current instruction to BlockStackEntry (#127482)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Support hasattr on BuiltinVariable (#127372)

anijain2305
authored and
pytorchmergebot
committed
[dynamo][compile-time] Cache the child guard managers (#127377)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 29, 2024

[dynamo] Support enum construction (#127364)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 28, 2024

[dynamo][fsdp] Skip Dynamo tracing of __getattr__ if its top-level frame (#127263)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 25, 2024

[dynamo][TensorVariable] Support "if param.grad_fn" usecase (#126960)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 23, 2024

[dynamo] Support __contains__ on obj.__dict__ (#126922)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 17, 2024

[dynamo] Sourceless builder - ordered dict and re.pattern (#126468)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 16, 2024

Commits on May 15, 2024

[dynamo][inline-inbuilt-nn-modules] Bug fix - Only unspecialized nn modules (#126303)

anijain2305
authored and
pytorchmergebot
committed
[dynamo] Detect monkeypatching on nn module forward method (#126203)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 14, 2024

[easy][dynamo] Use disable_dynamo for torch.manual_seed (#126192)

anijain2305
authored and
pytorchmergebot
committed
[reland][dynamo][disable] Move disable impl to its own __call__ method (#126191)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 11, 2024

[dynamo] Turn on guard_nn_modules (#125202)

anijain2305
authored and
pytorchmergebot
committed

Commits on May 9, 2024

[dynamo] Clear GenerationTracker on dynamo reset (#125855)

anijain2305
authored and
pytorchmergebot
committed
[dynamo][fsdp] Use Tensor match for FSDP modules (#125827)

anijain2305
authored and
pytorchmergebot
committed
[dynamo][logs] move recompilation reason within compile_id scope (#125805)

anijain2305
authored and
pytorchmergebot
committed

0