strisunshinewentingwang/pytorch

Pull Requests Commits

[MPS] Add `random_` overload (#98333)

malfet committed 1 year ago

29cde007

[c10d] delete lengths offset checks (#98368)

wanchaol committed 1 year ago

c9b1e099

[Dynamo] Fix bug of torch.is_floating_point & is_complex (#98393)

yanboliang committed 1 year ago

9c7b03d5

Add hacky example inputs to dynamo produced graph (#96561)

tugsbayasgalan committed 1 year ago

ebeaf8ad

Fix default argument of `parse_ir` stub (#98397)

take-cheeze committed 1 year ago

3d36f6f1

[pt2][inductor] retry add `triton.__verison__` as cache key, update cache layout (#98369)

nmacchioni committed 1 year ago

3c8e9e38

Python Dispatcher should respect FuncTorchBatchedDecomposition key (#98328)

zou3519 committed 1 year ago

f21a176c

Patch release process description (#98425)

atalman committed 1 year ago

78e991e5

Require sequence length in huggingface to be dynamic (#98335)

ezyang committed 1 year ago

37b91432

Require batch dimensions to be compiled dynamically (#98334)

ezyang committed 1 year ago

cf1bfca2

Don't error if we mark_dynamic without dynamic_shapes on (#98324)

ezyang committed 1 year ago

69f9bd23

Added ModuleInfos for Pooling ops (#98358)

mikaylagawarecki committed 1 year ago

2c6c7dee

[easy] Remove large LayerNorm sample input causing OOM from ModuleInfo (#98424)

mikaylagawarecki committed 1 year ago

3a0ad3c1

Add more debug logs to evaluate_expr (#98344)

ezyang committed 1 year ago

3ed66f94

remove //c10:headers (#98420)

mikey dagitses committed 1 year ago

f557402e

Make the Index Rounding Mode Consistent Between the 2D and 3D GridSample Nearest Neighbor Interpolations (#97000)

leimao committed 1 year ago

937ba248

[dtensor] add placement strategy and einsum strategy (#98227)

wanchaol committed 1 year ago

dcec2100

[pruning][core][feature] Implement convert for pruner (#97545)

jcaip committed 1 year ago

93063768

Add forward_prefetch flag to fully_shard (#98277)

rohan-varma committed 1 year ago

3657b37d

Clarify `Tensor.is_sparse` doc (#98408)

malfet committed 1 year ago

49c13025

Change daily aggregates upload job to use sum and occurence counter instead of averages (#98359)

PaliC committed 1 year ago

d1de5f5f

[spmd compile api] pre-flatten state container and pass the flattened state container to transforms (#98392)

yifuwang committed 1 year ago

762a81cb

[spmd compile API] add a (temporary) mechanism for overriding input tensors' placements (#98391)

yifuwang committed 1 year ago

37dbd5bf

[spmd expansion] support scalar_tensor (#98390)

yifuwang committed 1 year ago

970c08f9

[spmd expansion] speed up expansion by ~5x (#98389)

yifuwang committed 1 year ago

0830808d

[spmd expansion] support torch.ops.aten.sym_numel (#98388)

yifuwang committed 1 year ago

161f7c0b

Pattern matcher improvements (#97740)

jansel committed 1 year ago

3344d79e

Revert "[CUDA12] set_device change (#94864)"

pytorchmergebot committed 1 year ago

279ca5f9

Better Handling of Storage Cache (#98254)

eellison committed 1 year ago

981f9f04

Make sure we dealloc on recording, not just replay (#97440)

eellison committed 1 year ago

f1b901b0

Newer Older