strisunshinewentingwang/pytorch

Pull Requests Commits

dbr quant: refactor `get_quantized_op` to only use `seen_op_info` (#68342)

vkuzo committed 2 years ago

57472ec4

dbr quant: refactor `get_func_output_obs_type` to only use `seen_op_info` (#68341)

vkuzo committed 2 years ago

9cf4779e

dbr quant overhead[1/x]: remove expensive calls to named_modules (#68309)

vkuzo committed 2 years ago

f8b084c5

dbr quantization: inline scale and zp (#68251)

vkuzo committed 2 years ago

ed6ef0ee

barebones numeric suite for quantization with dynamic tracing (#67776)

vkuzo committed 2 years ago

ca499567

Strided masked softmin. (#68463)

pearu committed 2 years ago

d0eff8d8

[clone][sparse] Add `torch._C._sparse` namespace (#68672)

cpuhrsch committed 2 years ago

75955e4e

Implement topk with sort for some cases (#68632)

zasdfgbnm committed 2 years ago

95f4cd0b

Fix retry on connect failure decorator (#68600)

rohan-varma committed 2 years ago

e554d8b8

Make AOT compiler generic (#68637)

priyaramani committed 2 years ago

8e51381b

Rename shard_lengths to shard_sizes to be more inline with Tensor sizes. (#66464)

pritamdamania committed 2 years ago

c41d8290

Strided masked log_softmax. (#68461)

pearu committed 2 years ago

af564e73

Fix nanmedian result using more CUDA memory than necessary (#68591)

peterbell10 committed 2 years ago

578507cb

[fx2trt][easy] Replace all network.add_activation() call with helper function (#68676)

842974287 committed 2 years ago

6cca14d0

[torchelastic][1/n] Fix `caffe2.test.distributed.launcher.api_test` flaky tests (#68624)

aivanou committed 2 years ago

37edb748

[quant][graphmode][fx] Support input_quantized_idxs and output_quantized_idxs in the new convert (#68042)

jerryzh168 committed 2 years ago

a545a409

Remove doubly nested anonymous namespace (#68555)

peterbell10 committed 2 years ago

993b7a20

Add vectorized Jacobian and Hessian computation with forward AD (#67041)

soulitzer committed 2 years ago

5456d8c8

Add forward AD support for miscellanous operators (#67820)

soulitzer committed 2 years ago

7bb401a4

Add OpInfo test and fix a couple cases (#66294)

soulitzer committed 2 years ago

e358c49a

Add internal assert for tangent layout mismatch for view ops (#66293)

soulitzer committed 2 years ago

21d203b5

Address case when layout of tangent is not same as base (#66292)

soulitzer committed 2 years ago

2455cc2a

Support cuda 11.5: install magma for cuda in conda (#68665)

atalman committed 2 years ago

bbe2aae8

[reland] Fix flaky test_nccl_timeout (#68544)

rohan-varma committed 2 years ago

183dcdf5

Properly generate ToString methods where members are optional (#68560)

wconstab committed 2 years ago

Verified df6ced11

[quant][trt] Add support for torch.addmm in TensorRT (#67537)

jerryzh168 committed 2 years ago

875ba3dd

[SR] Add utility class to determine tensor ranges (#68284)

Mike Iovine committed 2 years ago

ee4cfaa2

[quant][graphmode][fx] Add support for weight and bias dtype in backend_config_dict (#68602)

jerryzh168 committed 2 years ago

a6d862c5

[ROCm] Use hipCUB/rocPRIM scan algorithms for large index support (#68487)

jaglinux committed 2 years ago

da4a95c7

Allow fuse unsqueeze cat sum with multiple input (#68650)

Shirong Wu committed 2 years ago

5880a2f1

Newer Older