strisunshinewentingwang/pytorch

Pull Requests Commits

[optim][asgd] group tensors in foreach to maximize perf

janeyx99 committed 1 year ago

c86c7fb7

Document hooks ordering behavior in the autograd note (#91667)

soulitzer committed 1 year ago

88366a90

Expose autograd.graph.Node as an abstract base class (#91475)

soulitzer committed 1 year ago

388b245d

[optim][adamw] default to foreach when CUDA + differentiable=False (#92306)

janeyx99 committed 1 year ago

0157e2ef

[onnx] Add mse_loss symbolic (#90717)

shubhambhokare1 committed 1 year ago

fcde6dbb

Update sdp_utils to check gradmode and subclassed tensors (#92323)

drisspg committed 1 year ago

40d6f2a0

Bypass filament2 for new pytorch random distribution method (#92190)

alyssawangqq committed 1 year ago

68f80420

Fix Windows cpu_profiling_allocator_test same pointer check flakiness (#92264)

huydhn committed 1 year ago

b31905c7

[torch.func] Add migration guide from functorch (#91811)

zou3519 committed 1 year ago

16f9d1bb

Revert "Improve `bsr @ strided` performance in `baddmm` for `bfloat16/half` with Triton kernels. (#88078)"

pytorchmergebot committed 1 year ago

89f1ad08

Improve `bsr @ strided` performance in `baddmm` for `bfloat16/half` with Triton kernels. (#88078)

nikitaved committed 1 year ago

7f256fff

Revert "Add sym_size/stride/numel/storage_offset to native_function.yaml (#91919)"

pytorchmergebot committed 1 year ago

befe8154

Revert "[FSDP] Do not clean FQNs even for `use_orig_params=True` (#91767)"

pytorchmergebot committed 1 year ago

88942a31

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

saitcakmak committed 1 year ago

0c8f4b58

[optim] abstract out _default_to_foreach_util (#92305)

janeyx99 committed 1 year ago

4fc796da

Revert "[fx] rewrite `FloorDiv` to match Python better (#90906)"

pytorchmergebot committed 1 year ago

5c9c39a8

Revert "[fx] fix type promotion in `binary_magic_impl` (#91376)"

pytorchmergebot committed 1 year ago

013afc5a

[pytorch] [compososable] make contract() pickle-able through functools wraps (#92120)

colin2328 committed 1 year ago

933cc67e

Run dynamo/test_dynamic_shapes serially (#92215)

huydhn committed 1 year ago

ea1007b8

Fix model accuracy issue caused by vectorized transpose (#92299)

jgong5 committed 1 year ago

2eaa7a25

Fix philox randn to follow standard normal distribution (#91945)

min-jean-cho committed 1 year ago

d29f0ba7

[FSDP] Do not clean FQNs even for `use_orig_params=True` (#91767)

awgu committed 1 year ago

d6f3265e

[FSDP][optim_state_dict][9/N] Rewrite the all-gather flow of optimizer state to support older GPUs (#91343)

fegin committed 1 year ago

1439cb03

Deprecate .mT,.T,.mH,.H on 0D tensors (#92143)

lezcano committed 1 year ago

46a81c8d

Perform first the decomposition and then the ATen function to catch in-place modifications (#92243)

lezcano committed 1 year ago

66e49862

Wrap a few more functions to ease their tracking during debugging (#92004)

lezcano committed 1 year ago

77b8aa6e

Add a test for decompositions that decomposes all the operations as much as possible (#87182)

lezcano committed 1 year ago

ea8b14f2

Assorted decomposition fixes (#87183)

lezcano committed 1 year ago

d162c8f9

Rewrite out-of-place decompositions in terms of out-of-place ops (#92003)

lezcano committed 1 year ago

da58f9eb

Check in some utility scripts for running dynamic shapes sweeps (#92256)

ezyang committed 1 year ago

1d47c593

Older