strisunshinewentingwang/pytorch

Pull Requests Commits

Speed up test_mps from 9min to 25s

albanD committed 2 years ago

dc2b2f09

Remove prints and add proper asserts

albanD committed 2 years ago

19ebdd7e

MPS: Fixes the as_strided_mps implementation for contiguous view operations (#78440)

kulinseth committed 2 years ago

f8160b11

MPS: Fix the memory growing issue and BERT_pytorch network crash fix. (#78006)

kulinseth committed 2 years ago

3e8119bf

[MPS] Fix `copy_kernel_mps` (#78428)

malfet committed 2 years ago

6660df9f

MPS: Eye op (#78408)

kulinseth committed 2 years ago

8b7e19a8

[mps] Do not use `malloc`/`free` in Indexing.mm (#78409)

malfet committed 2 years ago

98280132

MPS: Add adaptive max pool2d op (#78410)

kulinseth committed 2 years ago

53fc6dc3

MPS: add ranked tensors for addcmul ops instead of constants and update version_check (#78354)

kulinseth committed 2 years ago

52435c6b

Fix the MPS Heap volatility (#78230)

kulinseth committed 2 years ago

9a660613

Use random seed in normal_mps_out (#78010)

lhoenig committed 2 years ago

eef0ec54

Fix typo in testname (#78258)

malfet committed 2 years ago

0ffefea5

[MPS] Lazy initialize allocators (#78227)

malfet committed 2 years ago

7e12cfb2

[MPS] Add version check (#78192)

malfet committed 2 years ago

24b9bd43

Convert MPS Tensor data using MPSGraph API (#78092)

lhoenig committed 2 years ago

5342e760

[MPS] Fix torch.mps.is_available() (#78121)

malfet committed 2 years ago

08d70ab7

Add ignore for -Wunsupported-availability-guard

seemethere committed 2 years ago

207bde1e

Fix a few issues on assert/double error/legacy constructor (#77966)

albanD committed 2 years ago

51428a8f

[DataLoader] Apply sharding settings in dist when num_workers is 0 (#78967)

VitalyFedyunin committed 2 years ago

Verified c40f1845

[DataPipe] Adding functional API for FileLister (#78419) (#78948)

NivekT committed 2 years ago

Verified 8a5156a0

Make ShufflerDataPipe deterministic for persistent DL and distributed DL (#78765) (#78927)

ejguan committed 2 years ago

Verified 04d75d20

Avoid CPU Sync in SyncBatchNorm When Capturing CUDA Graphs (#78810)

mrshenli committed 2 years ago

Verified 2652da29

[chalf] warn once on creating a chalf tensor (#78245) (#78710)

kshitij12345 committed 2 years ago

Verified aa891188

[DataLoader] DataLoader now automatically apply sharding to DataPipes (#78762)

VitalyFedyunin committed 2 years ago

Verified 528710ec

[GHA] attempt to re-enable mac test workflows (#78000) (#78749)

atalman committed 2 years ago

Verified de53f70e

fix set item to scalar tensor missing gradient info (#78746)

yuguo68 committed 2 years ago

Verified 39ebb3e0

[DataPipe] Lazily generate exception message for performance (#78726)

NivekT committed 2 years ago

Verified fd3cc823

[docs][nn] conv: complex support note (#78351) (#78709)

kshitij12345 committed 2 years ago

Verified 5bb7c617

[DataLoader] Minor documentation improvement (#78548)

NivekT committed 2 years ago

Verified 8a627381

[ONNX] Fix typo when comparing DeviceObjType (#78085) (#78370)

atalman committed 2 years ago

Verified f56e16a7

Newer Older