strisunshinewentingwang/pytorch

Pull Requests Commits

print stderr for ghstack

clee2000 committed 1 year ago

2fecb64f

Add distributed dynamo benchmarking utils (#87419)

wconstab committed 1 year ago

7a4d91ca

Fix accuracy minifier (#87606)

ezyang committed 1 year ago

181b615b

sync AveragedModel buffers when use_buffers=False (#84054)

RangiLyu committed 1 year ago

512a3a48

[BE][einsum] add small comment explaining an invariant (#87264)

janeyx99 committed 1 year ago

1bcd63d5

[FSDP] `summon_full_params()` in computation stream (#86836)

awgu committed 1 year ago

a06e235e

[Quant][docs] Add README for BackendConfig (#86523)

andrewor14 committed 1 year ago

eafc910d

[FSDP][2/N] Remove `params_with_grad` (#87480)

awgu committed 1 year ago

084e7736

[FSDP][1/N] Rework `clip_grad_norm_()` and tests (#87479)

awgu committed 1 year ago

edac0d22

[FSDP][Docs] Clarify warnings to mention collectives (#87478)

awgu committed 1 year ago

3528b1fc

[FSDP] Rename streams (#86833)

awgu committed 1 year ago

573c8b6b

[FSDP] Use `reduce_scatter_tensor()` (#87240)

awgu committed 1 year ago

04ad0134

[xla hash update] update the pinned xla hash (#87590)

pytorchmergebot committed 1 year ago

cdb63a77

Simplify a few diagonal-related functions (#87180)

lezcano committed 1 year ago

faf9c47a

[PrimTorch] Add maker for *_copy variants of view functions (#87278)

lezcano committed 1 year ago

08c2314d

Improve readability of the extra message errors in assertEqual (#87202)

lezcano committed 1 year ago

5e4bcb04

Improvements for DDP Optimizer (#87549)

wconstab committed 1 year ago

233305a8

Fix 64bit indexing in `vol2col` (#87527)

eqy committed 1 year ago

4c8e1a98

[torch] Unify batch_box_cox implementations into perfkernels folder (#86569)

efiks committed 1 year ago

2e4c89eb

[Profiler] Regularize `AccumulateGrad` name (#86909)

Taylor Robie committed 1 year ago

0d2baed4

[Profiler][Trivial] Add Module cls and self bindings and type_caster macro (#86755)

Taylor Robie committed 1 year ago

5ec03fc1

[Profiler] Tensor IDs for Module and Optimizer variables (#86754)

Taylor Robie committed 1 year ago

b0e10292

[Profiler] Use parameter as key for optimizer state recording. (#86753)

Taylor Robie committed 1 year ago

be2d647e

Fix stupid N^2 naming behavior in FX and removed assert that slows things a lot sometimes (#87533)

Chillee committed 1 year ago

fc3beef5

[vision hash update] update the pinned vision hash (#87528)

pytorchmergebot committed 1 year ago

efdd43d5

Add xlogy and xlog1py references (#77712)

rdspring1 committed 1 year ago

9bb4926d

Fix meta for meta_fill_ (#87493)

SherlockNoMad committed 1 year ago

f3f1b447

[CI] Run all MacOS builds on MacOS-12 (#87496)

malfet committed 1 year ago

2f9fc160

[BE] Delete BUILD_SPLIT_CUDA option (#87502)

malfet committed 1 year ago

c28cdb53

Enable inductor CI for TIMM (#87462)

desertfire committed 1 year ago

f047dada

Older