8000 Commits · tombousso/pytorch · GitHub

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

Commits

Commits on Jun 8, 2024

Increase verbosity of FX graph dumps (#128042)

ezyang
authored and
pytorchmergebot
committed

Commits on Jun 6, 2024

Commits on Jun 4, 2024

Complete revamp of float/promotion sympy handling (#126905)

ezyang
authored and
pytorchmergebot
committed

Commits on Jun 3, 2024

Always simplify sympy expressions before printing. (#127543)

ezyang
authored and
pytorchmergebot
committed
Force Inductor output code to be dumped even if it fails to compile (#127700)

ezyang
authored and
pytorchmergebot
committed

Commits on Jun 2, 2024

Avoid reference cycle in inner closure (#127711)

ezyang
authored and
pytorchmergebot
committed

Commits on May 31, 2024

Unconditionally assign symbolic shapes as locals (#127486)

ezyang
authored and
pytorchmergebot
committed
Only register _inductor_test ops if not running with deploy (#127557)

ezyang
authored and
pytorchmergebot
committed
Add structured logging for tensor fakeification (#126879)

ezyang
authored and
pytorchmergebot
committed

Commits on May 29, 2024

Document the legacy constructor for Tensor (#122625)

ezyang
authored and
pytorchmergebot
committed

Commits on May 23, 2024

Don't add to replacements when guard is suppressed (#126210)

ezyang
authored and
pytorchmergebot
committed

Commits on May 22, 2024

Teach FakeTensor to fill in item_memo when converting scalar CPU tensor (#126245)

ezyang
authored and
pytorchmergebot
committed

Commits on May 21, 2024

Add guard_size_oblivious to vector_norm (#126772)

ezyang
authored and
pytorchmergebot
committed
Don't assume compare_arg is fx.Node (#126771)

ezyang
authored and
pytorchmergebot
committed
Add symbolic_shape_specialization structured trace (#126450)

ezyang
authored and
pytorchmergebot
committed

Commits on May 20, 2024

Fix bug for comptime.get_local for cells/closures (#126637)

ezyang
authored and
pytorchmergebot
committed

Commits on May 19, 2024

UFMT torch.utils._sympy.functions (#126553)

ezyang
authored and
pytorchmergebot
committed
Remove unnecessary implementations from MockHandler (#126511)

ezyang
authored and
pytorchmergebot
committed

Commits on May 17, 2024

Update ops handler documentation some more (#126480)

ezyang
authored and
pytorchmergebot
committed
55033ab
Add symbolic_shape_specialization structured trace (#126450)

ezyang
authored and
pytorchmergebot
committed

Commits on May 16, 2024

Make 'pytest test/inductor/test_memory_planning.py' work (#126397)

ezyang
authored and
pytorchmergebot
committed
Don't install inplace_methods on MockHandler, not needed (#126398)

ezyang
authored and
pytorchmergebot
committed
Generate runtime asserts when propagate real tensor is used (#126287)

ezyang
authored and
pytorchmergebot
committed

Commits on May 15, 2024

Make propagate_real_tensor more safe (#126281)

ezyang
authored and
pytorchmergebot
committed
Also remove compile_time_strobelight_meta frame when generating stack (#126289)

ezyang
authored and
pytorchmergebot
committed
Add VariableTracker.debug_repr (#126299)

ezyang
authored and
pytorchmergebot
committed
Improve Storage copy_ size mismatch error message (#126280)

ezyang
authored and
pytorchmergebot
committed
Parametrize test_dim_reduction (#126292)

ezyang
authored and
pytorchmergebot
committed
Beef up error message for pending assert failure (#126212)

ezyang
authored and
pytorchmergebot
committed
Don't assert about pending when we are peeking (#126239)

ezyang
authored and
pytorchmergebot
committed

Commits on May 14, 2024

Fanatically correct real tensor cloning for propagate_real_tensors (#126175)

ezyang
authored and
pytorchmergebot
committed
Move compute unbacked bindings call to track_tensor_tree (#126168)

ezyang
authored and
pytorchmergebot
committed

0