strisunshinewentingwang/pytorch

Pull Requests Commits

[ao][sparsity] Basic implementation of activation sparsifier (#80886)

macandro96 committed 2 years ago

f87d8c2f

Allow user to disable built-in fuser when using TorchDynamo (#81731)

wschin committed 2 years ago

f114468f

[Checkpoint] Fix autocasting (#81766)

rohan-varma committed 2 years ago

98cad3d3

[ao] Update DynamicStatic Detector to account for Conv (#81972)

vspenubarthi committed 2 years ago

75aab654

[ao] Fix to InputWeightEqualization detector to handle Conv groups (#81971)

vspenubarthi committed 2 years ago

0cacaf07

Revert "[_shard] make ShardedTensor be a Tensor and nn.Parameter (#79825)"

pytorchmergebot committed 2 years ago

f51cf774

Some fixes for FX pass with nvFuser backend (#81911)

SherlockNoMad committed 2 years ago

dc3c1ade

Update the pull request template (#81991)

ZainRizvi committed 2 years ago

89a66800

[FSDP] Move `_post_backward_called` to `_init_param_attributes` (#81243)

awgu committed 2 years ago

9c94e10b

[FSDP] Clean up `_lazy_init()` (#80185)

awgu committed 2 years ago

777ff539

[Easy][FSDP] Add `zero_grad()` to unit test train loop (#80087)

awgu committed 2 years ago

4bf33ab8

[FSDP] Remove `self.numel_padded_per_param` (unused) (#80002)

awgu committed 2 years ago

3059b137

[FSDP] Move tensor sharding logic to `FlatParamHandle` (#80000)

awgu committed 2 years ago

790b1229

[FSDP] Deduplicate `_orig_size` and `_unsharded_size` (#79984)

awgu committed 2 years ago

b069120d

[FSDP] Introduce `FlatParamHandle` (#79652)

awgu committed 2 years ago

be656f55

get rid of push_torch_{dispatch, function}_mode (#78215)

samdow committed 2 years ago

2ac24675

Revert "Add more functorch shards to PR CI (#81919)"

pytorchmergebot committed 2 years ago

3bd08e34

Revert "Add python stack tracing option on on-demand flow (#80919)"

pytorchmergebot committed 2 years ago

6381cd61

Lazy shape symbolic const (#81718)

miladm committed 2 years ago

1a632920

Adding maximize to ASGD (#81875)

zaxtax committed 2 years ago

50c655d5

[_shard] make ShardedTensor be a Tensor and nn.Parameter (#79825)

wanchaol committed 2 years ago

9c32439a

[rpc/distributed] eliminate code duplication in distributed/rendezvou… (#81577)

Bill Darrow committed 2 years ago

38988a8d

[vulkan] implement quantized depthwise conv2d (#81497)

psavla2 committed 2 years ago

9f9cff2b

[vulkan] implement quantized pointwise conv2d (#81496)

psavla2 committed 2 years ago

a86aa478

[vulkan] implemented quantized reg conv2d (#81495)

psavla2 committed 2 years ago

0481cb81

[vulkan] implement quantized add (#81494)

psavla2 committed 2 years ago

054d4c12

[ONNX] Deprecation utility (#81901)

justinchuby committed 2 years ago

74c6a754

Add `scalarType` to JitType pyi (#81112)

justinchuby committed 2 years ago

1bdb1a0a

cudagraphs dynamo backend (#80566)

ezyang committed 2 years ago

3c2c2cc9

fix LTC usage of torch.tensor ctr, add test (#81928)

bdhirsh committed 2 years ago

38023ff3

Newer Older