strisunshinewentingwang/pytorch

Pull Requests Commits

[ONNX] Replace decomposeLinear pre process pass with a symbolic (#53077) (#54866)

KsenijaS committed 3 years ago

cb0cee4a

[ONNX] Fix if output shape mismatch error & Fix graph input directly used as output (#53219) (#54865)

BowenBao committed 3 years ago

849dcb8b

[ONNX] Support primitive type input/outputs and attributes (#53550) (#54864)

neginraoof committed 3 years ago

cd9dd653

[ONNX] Improve index_put symbolic to handle singular Bool updates (#53690) (#54863)

shubhambhokare1 committed 3 years ago

ce48b140

Allow `std=0` in `torch.normal`, and error if `std<0` (#51317)

kurtamohler committed 3 years ago

6c235ef2

Revert D27408378: [quant][graphmode][fx][refactor] Factor out insert_observers_for_model to a separate function

Mike Ruberry committed 3 years ago

15f04e34

Added OpInfo gradcheck_wrapper to replace output_func (#54914)

heitorschueroff committed 3 years ago

8b8c4096

Language Ref for Statements in Torchscript (#52847)

nikithamalgi committed 3 years ago

790b69e0

[quant][graphmode][fx][refactor] Factor out insert_observers_for_model to a separate function (#54733)

jerryzh168 committed 3 years ago

c445f4ee

[quant][graphmode][fx] Separate handling Copy operator to a helper function (#54644)

jerryzh168 committed 3 years ago

c57541ce

[quant][fx][graphmode][refactor] Change activation_post_process_map to track the observer name instead (#54643)

jerryzh168 committed 3 years ago

c0d6dbdc

[quant][graphmode][refactor] Remove reduandent code (#54073)

jerryzh168 committed 3 years ago

c2adedf6

[quant][graphmode][fx] Add support for one value being quantized with different qconfigs (#53586)

jerryzh168 committed 3 years ago

55544cb1

Revert D27469727: [pytorch][PR] [android] fbjni from prefab dependency 0.2.2

Ailing Zhang committed 3 years ago

eb52e364

Move shape prop inside acc_tracer (#55091)

jfix71 committed 3 years ago

c85d3f50

fix boto3 resource not close (#55082)

Rong Rong (AI Infra) committed 3 years ago

0d80f378

Refactor get numerical jacobian (#54092)

soulitzer committed 3 years ago

f2903967

Remove use_c10_dispatcher option (#54969)

wenleix committed 3 years ago

70af5db7

[Refactoring] make transformations return whether graph is modified (#54777)

Peng Wu committed 3 years ago

908a74e8

[Futures] Bump log verbosity when ignoring cb errors in python future. (#54476)

rohan-varma committed 3 years ago

a37fbf9b

[Futures] enhance error handling in then() (#54475)

rohan-varma committed 3 years ago

28daa6b7

various overhead improvements to cuda addmm (#55026)

ngimel committed 3 years ago

63c70ae0

Clarify tools/test_history.py output for re-runs (#55106)

samestep committed 3 years ago

8eb9a934

[Pytorch] Expose ops present in dispatcher (#54791)

JacobSzwejbka committed 3 years ago

2726de01

Handle 1D input for xnnpack::linear (#54986)

Akshit Khurana committed 3 years ago

5c396337

Simplify convolution double backward gradInput formulas (#54840)

ngimel committed 3 years ago

fb1c193e

[ROCM] enable miopen for rnn f16 (#52475)

micmelesse committed 3 years ago

26c1e2ee

[android] bump nigtlies version to 1.9.0 (#55076)

IvanKobzarev committed 3 years ago

7f873588

[FX] Add a metadata dict to Node and switch shapeprop to use that (#54926)

James Reed committed 3 years ago

bcb45831

Adding workflow to auto-label PRs with ROCm (#54989)

janeyx99 committed 3 years ago

b64d7756

Newer Older