strisunshinewentingwang/pytorch

Pull Requests Commits

use irange for loops 8 (#66743)

r-barnes committed 2 years ago

17e52004

[BE] rewrite ProcessGroupNCCLTest to be MultiProcess (#67705)

wanchaol committed 2 years ago

ff3fc372

ns for fx: fix bug in graph matcher (#69238)

vkuzo committed 2 years ago

5c816520

Add functorch TLS to ATen/ThreadLocalState (#69181)

zou3519 committed 2 years ago

698c35e7

functionalization: update is_aliased() logic (#68881)

bdhirsh committed 2 years ago

0de7a618

[quant][embdding qat] Add FX support for QAT EmbeddingBag (#68121)

b-koopman committed 2 years ago

4484c045

Do not modify type map from getCustomClassTypeImpl() (#69261)

malfet committed 2 years ago

78ab3cde

Fix crash in `checkCustomClassType` if arg is null (#69259)

malfet committed 2 years ago

113684cf

Add efficient zero tensors (#64837)

anjali411 committed 2 years ago

668574af

[quant][embedding qat] Support Embedding QAT via FX API (#68296)

b-koopman committed 2 years ago

abda069c

[quant][embedding qat] Fix bug enforcing quant_min <= zero_point <= quant_max for float zeropoint (#68852)

b-koopman committed 2 years ago

3157371b

Add Lazy Tensor codegen infra (#69020)

wconstab committed 2 years ago

397183f4

Follow the undefined Tensor <-> None rule better in torch dispatch (#67793)

albanD committed 2 years ago

28c51996

[DataPipe] Adding BatcherMapDataPipe (#68197)

NivekT committed 2 years ago

0465f64b

Revert D32010095: [pytorch][PR] Add ability for a mobile::Module to save as flatbuffer

albanD committed 2 years ago

00ebbd5e

[Static Runtime] Skip ProcessedNode:: verify_no_memory_overlap() for out variants (#68639)

Hao Lu committed 2 years ago

ed3b73fd

[shard] add back init_from_local_shard_and_global_metadata API (#69226)

wanchaol committed 2 years ago

c60232d8

support pure fp16 training in FSDP (#68417)

zhaojuanmao committed 2 years ago

12621c3a

Add ability for a mobile::Module to save as flatbuffer (#67351)

qihqi committed 2 years ago

41d35dc2

[JIT] Compute input sym shapes in partial eval graph (#68281)

Elias Ellison committed 2 years ago

40fb28ea

[DataPipe] Simplify BatcherIterDataPipe by removing 'unbatch_level' argument and functionality (#68594)

NivekT committed 2 years ago

d8a44270

[deploy] docs (#69251)

suo committed 2 years ago

ad182479

Back out "[CUDA Pinned Memory] Event recording with non-blocking copies should track the storage context, not the tensor data pointer" (#69193)

dstaay-fb committed 2 years ago

cbe0a38d

Back out "[CUDA Pinned Memory] Alternative implementation of pinned memory allocator focusing on multi-threaded scalability" (#69191)

dstaay-fb committed 2 years ago

929f2a75

Strided masked var. (#68738)

pearu committed 2 years ago

370d0afc

[Pytorch Edge] Update Black Box Api with operator versioning (#68678)

JacobSzwejbka committed 2 years ago

291e56ed

[Operator Versioning][Edge] Add old models and unittest (#67726)

cccclai committed 2 years ago

b9738e92

RegisterDispatchKey.cpp: remove redundant code (#68983)

guoyejun committed 2 years ago

124bb6a1

[torch][distributed] Check for file existence before invoking cleanup logic in FileStore destructor (#68603)

aivanou committed 2 years ago

fced51ea

fixing layer_norm cuda bug (#69210)

jjsjann123 committed 2 years ago

3c1e2ff9

Newer Older