strisunshinewentingwang/pytorch

Pull Requests Commits

[functorch] Update README.md

Chillee committed 2 years ago

236d2f20

[functorch] Added a quick benchmark

zou3519 committed 2 years ago

046453c6

[functorch] Added citation section

zou3519 committed 2 years ago

8d127816

[functorch] Fix parametrized testing

zou3519 committed 2 years ago

57f48cc6

[functorch] Fix grad wrapper behavior when wrapping Python tensor

zou3519 committed 2 years ago

fea6254d

[functorch] removed comment that previous commit fixed :)

Chillee committed 2 years ago

e836870e

[functorch] ported random registrations to boxedfallback

Chillee committed 2 years ago

d71bb374

[functorch] Added errors for dynamic ops

Chillee committed 2 years ago

b28268c8

[functorch] Batch rule for resize_ when batch dim is at front

zou3519 committed 2 years ago

899fea5e

[functorch] Added dropout (tested by opinfo to be upstreamed)

Chillee committed 2 years ago

0bba49c8

[functorch] Added softmax batching rule (tested through local OpInfo, to be upstreamed)

Chillee committed 2 years ago

a598ed6b

[functorch] Added dist decomposition

Chillee committed 2 years ago

abd8c710

[functorch] moved some decompositions from batchingregistrations.cpp out

Chillee committed 2 years ago

23fc2e0f

[functorch] added remove-inplace pass to nnc_jit (kinda works, but won't work with aliasing)

Chillee committed 2 years ago

71042b1b

[functorch] Update dependency (pytorch/functorch#58)

zou3519 committed 2 years ago

abc520f8

[functorch] (Grad)TensorWrapper sometimes has storage (and data_ptr) (pytorch/functorch#65)

zou3519 committed 2 years ago

a9ac8e81

[functorch] Install expecttest as CI step (pytorch/functorch#66)

zou3519 committed 2 years ago

26503f21

[functorch] added ability for module attrs to be directly referenced in output

Chillee committed 2 years ago

74f77319

[functorch] Added binary_cross_entropy lowerings

Chillee committed 2 years ago

f10845ee

[functorch] fixed device issues in tests

Chillee committed 2 years ago

379ae35e

[functorch] add no-dim for batchrulesreduce and move trace from old batching rule to new decomposition

Chillee committed 2 years ago

f561f0f6

[functorch] Removed no-dim rules for batchrulespooling and batchrulesviews

Chillee committed 2 years ago

0a8ef33c

[functorch] removed no bdim batching rules from batchrulesbinaryops

Chillee committed 2 years ago

8f06e915

[functorch] removed no bdim cases from batchruleslinearalgebra

Chillee committed 2 years ago

cb286b9b

[functorch] Added a test for case where we would previously have dispatched to a batching rule despite having no bdims

Chillee committed 2 years ago

8cd80e0b

[functorch] Selectively enable dispatch on kBatchedKey (pytorch/functorch#63)

zou3519 committed 2 years ago

446d0b4e

[functorch] Added einsum batching rule

Chillee committed 2 years ago

bbfdcdbd

[functorch] updated opinfo DB and added meshgrid batching rule

Chillee committed 2 years ago

495112f5

[functorch] Exclude failing tests, add xfail test for torch.tensor

zou3519 committed 2 years ago

f6667347

[functorch] Add flip batching rule

Chillee committed 2 years ago

c02cc07c

Newer Older