strisunshinewentingwang/pytorch

Pull Requests Commits

add additional stream priority for cuda streams (#101956)

ngimel committed 1 year ago

5da497ca

update tf32 thresholds in nn/test_convolution.py (#102015)

Fuzzkatt committed 1 year ago

f8896b7b

No need to run non-CUDA jobs in memory leak check mode (#102188)

huydhn committed 1 year ago

dedcf8f7

[inductor][decomp] Add aten._unsafe_index_put for unchecked indexing (#101812)

peterbell10 committed 1 year ago

ce420107

Populate all args with fake tensor value (#102129)

tugsbayasgalan committed 1 year ago

dbf6912b

Revert "Support resize on meta storage (#101988)"

pytorchmergebot committed 1 year ago

210fc28d

[distributed] implemented find_all_descendants (#102138)

moonbucks committed 1 year ago

06f656c5

[dynamo][higher order op] Support nn.Module calls (#102022)

anijain2305 committed 1 year ago

5d6810a4

[dynamo] Add astunparse dependency (#102120)

desertfire committed 1 year ago

e6af31a5

Segmentation fault in flatbuffers when parsing malformed modules (#95221)

apach301 committed 1 year ago

e6fc7d81

torch.sparse.softmax: allow negative dim (#102172)

nikitaved committed 1 year ago

2e2a7467

Add quantization lowering for nn.PixelShuffle and nn.PixelUnshuffle (#101926)

dulinriley committed 1 year ago

424c930f

add ignored_states to FSDP/fully_shard (#102056)

zhaojuanmao committed 1 year ago

956bd038

Revert "Merge type stubs for torch.nn.parallel (#101528)"

pytorchmergebot committed 1 year ago

023bc30b

[spmd] Enable data parallel to work with non 0 batch dim (#100073)

wanchaol committed 1 year ago

d316a2dd

[spmd] add more decomp and fix a sharding bug (#100938)

wanchaol committed 1 year ago

d3788370

[spmd] Improve activation handling, factory ops and batch dim reduction (#100853)

wanchaol committed 1 year ago

dd1f2952

[spmd] enhance batch dim analysis of data parallel (#100852)

wanchaol committed 1 year ago

4d55ea85

[spmd] by default average gradients for nccl backend (#99964)

wanchaol committed 1 year ago

b2eaba6b

[spmd] add option to preserve node types (#100072)

wanchaol committed 1 year ago

942cd12d

No cpp + step current (#102001)

clee2000 committed 1 year ago

2232cce6

Unbind Cat pattern (#101767)

devashishshankar committed 1 year ago

fcf812c3

Merge type stubs for torch.nn.parallel (#101528)

ringohoffman committed 1 year ago

6cabc105

Replace unsqueeze transform with stack (#101766)

devashishshankar committed 1 year ago

fdd28399

Replace split-squeeze pattern (#101765)

devashishshankar committed 1 year ago

c0d0a9f7

[ONNX] Introduce FX-ONNX dispatcher (#100660)

titaiwangms committed 1 year ago

0eb4f072

Refactor normalize passes to use @register_graph_pattern (#101764)

devashishshankar committed 1 year ago

47b41364

[CUDA][DLPack] Handle legacy default streams for DLPack conversion (#101318)

eqy committed 1 year ago

66f6e0e6

[quant][pt2e][be] Move annotate helper function to quantizer/utils.py (#102127)

jerryzh168 committed 1 year ago

3baa67ca

[inductor] Move two cpu tests to test_cpu_repro.py (#101887)

desertfire committed 1 year ago

5f0463a6

Newer Older