strisunshinewentingwang/pytorch

Pull Requests Commits

[TorchTidy] Add pattern to detect if single-tensor implementation optimizer is used (#81733)

davidchencsl committed 2 years ago

aeb97d95

[Profiler] Add speedup estimate for FP32 pattern and Extra CUDA Copy Pattern (#81501)

davidchencsl committed 2 years ago

64c6387c

[FSDP] Update `ShardingStrategy` and `_free_full_params()` docs (#80894)

awgu committed 2 years ago

10a47c53

add all missing constraint for HFmodel (#81714)

migeed-z committed 2 years ago

2c57b27c

[ao][sparsity] Serialization support (#80890)

macandro96 committed 2 years ago

1ba63e5a

[ao][sparsity] Implementation of squash_mask() (#80889)

macandro96 committed 2 years ago

aa234479

[ao][sparsity] Implementation of step() and update_mask() (#80888)

macandro96 committed 2 years ago

6b3bf3d6

[ao][sparsity] Implementation of register_layer() and get_mask() (#80887)

macandro96 committed 2 years ago

5fe3a166

[ONNX] Remove duplicated test run (#81146)

BowenBao committed 2 years ago

4c0000a9

[ONNX] Fix bug using std::copy_if (#80999)

BowenBao committed 2 years ago

df4226b4

[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

Newer Older