strisunshinewentingwang/pytorch

Pull Requests Commits

[optim][asgd] default to foreach when CUDA + differentiable=False (#92724)

janeyx99 committed 1 year ago

9f356568

[optim][adamax] default to foreach when CUDA + differentiable=False (#92723)

janeyx99 committed 1 year ago

30bda6b1

[optim][adagrad] default to foreach when CUDA + differentiable=False (#92716)

janeyx99 committed 1 year ago

9b4a7784

Print aot graphs if user specifies aot graph env vars (#92720)

mlazos committed 1 year ago

6f1727b2

Use SymBool for is_contiguous computation (#92229)

ezyang committed 1 year ago

c0fe41f9

[vision hash update] update the pinned vision hash (#92732)

pytorchmergebot committed 1 year ago

011df663

[functorch] add is_any_true (#92686)

kshitij12345 committed 1 year ago

d2728bb6

[pt2.0/inductor] Fix race in cache dir across ranks on the same host (#92664)

xw285cornell committed 1 year ago

e6a8267c

[BE][CI] rename .jenkins to .ci, add symlink (#92621)

janeyx99 committed 1 year ago

8972a9fe

Revert "Update Module.__setattr__ to respect property setters (#92044)"

pytorchmergebot committed 1 year ago

09eb4c2a

remove useless clang-tidy suppression (#92287)

cyyever committed 1 year ago

85851b1e

Add periodic job to test aot_eager on benchmarks suite. (#92695)

ezyang committed 1 year ago

5489b323

Update aot_eager CI failures (#92696)

ezyang committed 1 year ago

9ad0aca6

Refactor test_inductor_benchmark into test_single_dynamo_benchmark helper (#92665)

ezyang committed 1 year ago

1bf51201

Add a warning about performance cost of set_default_device (#92703)

ezyang committed 1 year ago

85a1f022

Implement SymBool (#92149)

ezyang committed 1 year ago

5c6f5439

Dispatch the auxiliary frobenius_norm and nuclear_norm to better implementations and deprecate them (#81763)

lezcano committed 1 year ago

34e8eb22

[cublas][cublasLt] Fall back to unfused `addmm` for 2-byte-aligned inputs (#92201)

eqy committed 1 year ago

1af40d51

[quant][fx][pt2e][be] Store node_name_to_target_dtype to node.meta["target_dtype_info"] (#92574)

jerryzh168 committed 1 year ago

a74c8df7

[optim][foreach] Do NOT inplace modify gradients (#92706)

janeyx99 committed 1 year ago

de0375e7

[optim][NAdam] Fix discrepancy between mt vs st impl (#92699)

janeyx99 committed 1 year ago

2b885e1f

fix the formatting of runtime error msg in prims _cat_meta (#92124)

dshi7 committed 1 year ago

896b6d87

Shard mac to 3 (#91277)

clee2000 committed 1 year ago

703265e5

Don't allow recomputing a node that *must* be materialized in the backwards pass (#90896)

Chillee committed 1 year ago

d6c3468f

Fix GroupNorm backward prop on CUDA (#92671)

malfet committed 1 year ago

97b7e4cd

[CUDA][CUBLAS][BFloat16] Tenatively disable reduced precision reductions for some matmul tests (#92599)

eqy committed 1 year ago

8c0289a6

[inductor] Lower torch.exp2 and use it for torch.pow(2, x) (#92632)

peterbell10 committed 1 year ago

56440594

Add GHA side support for ciflow/inductor-perf-test-nightly (#92693)

ezyang committed 1 year ago

5a134440

Create autograd Function for aot_autograd backward only when needed (#92688)

soulitzer committed 1 year ago

a3efa9d7

[PT-D][checkpoint] Resolve no such file or directory issue when checkpointing on multi hosts (#92553)

wz337 committed 1 year ago

eee2869e

Newer Older