strisunshinewentingwang/pytorch

Pull Requests Commits

Fix round robin sharding (#121022)

Catherine Lee committed 2 years ago

68015953

Update NCCL submodule to v2.20.5 (#121635)

Skylion007 committed 2 years ago

e2ac2dc1

fix synchronization behavior for copies with type change (#121341)

Natalia Gimelshein committed 2 years ago

89add711

Fix lower precision check for MKLDNN on Windows (#121618)

CaoE committed 2 years ago

03717430

Add NEON accelerated torch.mv kernel (#119992)

malfet committed 2 years ago

e2900461

CI sanity check test for env vars (#120519)

Catherine Lee committed 2 years ago

fac06a12

Add support to save safetensors checkpoint directly into onnx (#121001)

Thiago Crepaldi committed 2 years ago

6c11d3ce

add __repr__ function to FunctionSchema for Python (#121484)

fffrog committed 2 years ago

485f8ebc

Upgrade submodule onednn to v3.3.5 (#120767)

Xia-Weiwen committed 2 years ago

d1510e01

[dtensor][debug] force visualize_sharding not to print for empty tensors (#121217)

XilunWu committed 2 years ago

605c0a28

[dtensor][debug] visualize_sharding skip if the current rank is not in mesh (#121382)

XilunWu committed 2 years ago

3a5ab17b

[dtensor][debug] visualize_sharding only compute offset on the first rank in mesh (#121385)

XilunWu committed 2 years ago

b383123e

Fix `get_rank` under a non-default group. (#120481)

KungYork committed 2 years ago

9c50ecc8

[dynamo] Fix support for nn.Parameter constructor (part 1) (#120163)

jansel committed 2 years ago

7cc476ea

[dynamo] Support _unsafe_set_version_counter (#121086)

jansel committed 2 years ago

32488b06

[Dynamo] Fix function overrides (#120885)

StellarrZ committed 2 years ago

7a4e4511

split predispatch pass into multiple passes (#121592)

kflu committed 2 years ago

f11f2b0d

relax assertion on fake shape (#121599)

avikchaudhuri committed 2 years ago

13e8181b

[Export] Fix bug removing node from wrong graph (#121574)

oulgen committed 2 years ago

660ec3d3

[IntraNodeComm] fix a hybridCubeMeshAllReduceKernel breakage caused by a recent refactor (#121575)

Yifu Wang committed 2 years ago

41286f15

[DeviceMesh] Add support for nD slicing (#119752)

wz337 committed 2 years ago

60cd2a43

[inductor] Log triton kernel source and metadata on failure (#120494)

bertmaher committed 2 years ago

e90cddb0

[inductor] Changes to support newer triton pin (#121267)

peterbell10 committed 2 years ago

168a04e7

[inductor] Refactor common triton imports into one function (#121438)

peterbell10 committed 2 years ago

459c5bca

Remove opmath cast for im2col decomp (#121363)

BowenBao committed 2 years ago

8c96b436

[dynamo] support rewriting dist.all_reduce with explicitly specified reduce op (#120181)

Yifu Wang committed 2 years ago

71d02026

Revert "Add CUTLASS kernel as choice for _int_mm() Inductor autotuning (#119685)"

pytorchmergebot committed 2 years ago

cf974237

[profiler] Fix recorded profiler step number (#121127)

fwenguang committed 2 years ago

761783a4

[dtensor] add async_op option to redistribute and some refactor (#121477)

wanchaol committed 2 years ago

242e03ba

[quant] Add error check for input_edge annotation (#121536)

jerryzh168 committed 2 years ago

a6a67da3

Newer Older