strisunshinewentingwang/pytorch

Pull Requests Commits

[pt2] support vmap (#101707)

kshitij12345 committed 1 year ago

cce2c52b

Don't suppress ModuleNotFoundError if the failure is for an unrelated module (#106807)

ezyang committed 1 year ago

c379d628

[quant][pt2e] Add reference representation for quantize_per_channel and dequantize_per_channel (#105783)

jerryzh168 committed 1 year ago

69ecad6f

expandable_segments fix possible assert (#106818)

zdevito committed 1 year ago

c14cf312

[CI] Fix sccaching of nvcc builds (#106811)

malfet committed 1 year ago

44448754

[pytorch][ao] force weight observer/fake_quant to be on the same device as the weight tensor (#106755)

jiaxuzhu92 committed 1 year ago

9e35df4a

[DCP] Modify tensor saving logic in DCP (#106415)

eddogola committed 1 year ago

cbcd9083

Remove dynamo+nvfuser (#105789)

IvanYashchuk committed 1 year ago

c913f385

Enable mypy check for torch/_inductor/codegen/common.py (#106199)

ashok-kumar-kannan committed 1 year ago

32d8de23

[ONNX] Turn on batch norm related unittest (#105769)

BowenBao committed 1 year ago

2a138d7f

[pytorch] Disable CUDA sync events by default (#106723)

briancoutinho committed 1 year ago

3c52c6fd

[exir] Update exir.pass_base to use export.pass_base (#106647)

angelayi committed 1 year ago

d4bc2719

Add missing quantize per tensor vulkan backend function (#106641)

shubhraprakash1 committed 1 year ago

2764ead4

[export] refactor and add same_signature flag to dynamo.export (#106569)

ydwu4 committed 1 year ago

3a300ed8

add a debug api to extract cache entry from code (#106673)

youkaichao committed 1 year ago

bd3b6f1a

Back out "Reland "Make adding buffers more like adding parameters (#104069)" (#106224)" (#106743)

rocketerfb committed 1 year ago

bc88028e

Revert "Remove dynamo+nvfuser (#105789)"

pytorchmergebot committed 1 year ago

891bb259

[custom_ops] extend impl_abstract to work with existing torch.library ops (#106088)

zou3519 committed 1 year ago

16b68738

[custom_ops] make custom_ops.impl work on existing operators (#106076)

zou3519 committed 1 year ago

cebff39f

[custom_ops] Block overload names (#106075)

zou3519 committed 1 year ago

60a4ac30

Remove dynamo+nvfuser (#105789)

IvanYashchuk committed 1 year ago

6030151d

[core][pruning][feature] cuSPARSELt kernels and ops (#102133)

jcaip committed 1 year ago

ad22f0ff

[CUDA][CUDA Graphs] Fix potential race with autograd thread during a graph capture 2 (#106570)

eqy committed 1 year ago

03c93217

Enable dynamic shapes of torch.nn.Parameter (#105855)

bohnstingl committed 1 year ago

cc21fa75

Fix test_div in caffe2/caffe2/python:hypothesis_test (#106694)

MaigoAkisame committed 1 year ago

0d57e870

[MPS] Introduce torch.mps.Event() APIs (#102121)

razarmehr committed 1 year ago

cdfd0ea1

[NCCL][AVOID_RECORD_STREAMS] Initialize `stashed_for_allocator_safety_` in `endCoalescing` if `TORCH_NCCL_AVOID_RECORD_STREAMS=1` (#106166)

eqy committed 1 year ago

5f551133

Add 0-dim `Tensor` overload to `_foreach_mul` (#106677)

crcrpar committed 1 year ago

9e4e0ecd

sd flaky on cpu skip (#106726)

msaroufim committed 1 year ago

90c264c2

inductor: remove conv_bn folding from pre_grad pass (#106686)

XiaobingSuper committed 1 year ago

2a164579

Newer Older