strisunshinewentingwang/pytorch

Pull Requests Commits

make triton uses the wheel script now (#97995)

ezyang committed 2 years ago

f2e6b083

Skip test_batch_norm in test_jit_fuser_te for asan (#98016)

clee2000 committed 2 years ago

1f85390e

[vmap] Fix index_select support when dim is negative (#97916)

zou3519 committed 2 years ago

7bb5fb3c

Revert "Disable dynamo tracing torchrec.distributed (#97824)"

pytorchmergebot committed 2 years ago

7868e4b4

Fix module backward pre-hooks to actually update gradient (#97983)

soulitzer committed 2 years ago

ee1c539e

[dynamo 3.11] fix push null timing in resume functions (#96504)

williamwen42 committed 2 years ago

06d677f4

[dynamo 3.11] properly determine cell/freevar index in bytecode_transformation.py (#96503)

williamwen42 committed 2 years ago

5b6e4c48

[dynamo 3.11] properly copy free/cell vars in eval_frame.c (#96501)

williamwen42 committed 2 years ago

ba52268d

[inductor] fix TritonTemplateCaller.__str__ (#97578)

shunting314 committed 2 years ago

c681c52e

[dynamo 3.11] fix eval_frame.c debug prints for 3.11 (#96500)

williamwen42 committed 2 years ago

c905251f

fix functional collective to not generate getattr node (#97924)

wanchaol committed 2 years ago

848bf810

[CUDA][CUDA 11] Remove more CUDA 11 version checks (#92934)

eqy committed 2 years ago

2fddcf0f

[inductor] test codegen with dynamic shapes (#96934)

nkaretnikov committed 2 years ago

90f69cad

distinguish mutability of StorageImpl::data_ptr() member (#97651)

mikey dagitses committed 2 years ago

da28af32

set num_warps to at least 4 (#97950)

Natalia Gimelshein committed 2 years ago

35090b86

Fix TorchScript support in `as_nested_tensor` (#97960)

rusty1s committed 2 years ago

19706356

Compare `len(fw_derivatives)` with 0 w/o using `not` (#97953)

crcrpar committed 2 years ago

b235e1f7

Run the benchmark suite with dynamic batch only (#97912)

ezyang committed 2 years ago

97fc8ea5

Move TestIndexingSimplification to its own file (#97941)

jansel committed 2 years ago

4cce6075

Fix strip_function_call in GuardBuilder (#97810)

lantiankaikai committed 2 years ago

94bae36a

[fix] take : backward batching rule (#95772)

kshitij12345 committed 2 years ago

ffd76d11

Retry at test file level (#97506)

clee2000 committed 2 years ago

7d5d5beb

[dynamo 3.11] Refactor create_instruction (#96499)

williamwen42 committed 2 years ago

24a5d006

Revisit `torch._six.string_classes` removal (#94709) (#97863)

XuehaiPan committed 2 years ago

e6888697

Enable adam foreach in full train step tracing (#97897)

mrshenli committed 2 years ago

9ec6fdb2

[functorch] .data should not work for grad, jvp, vjp (#94817)

kshitij12345 committed 2 years ago

19dcf55a

Add unstable workflow to upload test stats (#97918)

clee2000 committed 2 years ago

96dbca69

Corrected batch norm docs with the exact computations of the standard deviation (#97974)

zaccharieramzi committed 2 years ago

65e8c149

[minifier] cuda.synchronize to better detect IMA (#97962)

anijain2305 committed 2 years ago

cdb32dad

NT: Refactor for lazy computation of opt_sizes (#97895)

jbschlosser committed 2 years ago

0e4ddc2b

Newer Older