strisunshinewentingwang/pytorch

Pull Requests Commits

Temporarily move ROCm to unstable (#99579)

kit1980 committed 1 year ago

d06624d3

Add an expect test for test_save_graph_repro (#99538)

ezyang committed 1 year ago

805a6dc8

[quant][pt2e][refactor] Move the annotation for observer sharing ops into separate util (#99384)

jerryzh168 committed 1 year ago

36acad58

increment generation in run only context (#99099)

eellison committed 1 year ago

1b9edb68

Adding vmap support for special bessel functions (#99543)

SiddharthIVEX committed 1 year ago

c660db20

Remove a couple of additional places where we would construct tensors - aliases of params, inputs (#98950)

eellison committed 1 year ago

19788002

Add TorchXLA option to benchmark runner (#99505)

jansel committed 1 year ago

3233450d

Make HAS_CPU boolean lazy, speed up import time (#99537)

ezyang committed 1 year ago

6026caed

Don't eagerly initialize CUDA when importing common_cuda (#99536)

ezyang committed 1 year ago

cf354a04

Package `torch.fx` type hints (#99541)

malfet committed 1 year ago

32cd05ae

suggest constraints to specify for export based on generated shape guards (#98463)

avikchaudhuri committed 1 year ago

f6f35135

Support module dict iter (#99503)

voznesenskym committed 1 year ago

04f7a2a5

[cond] error on closed over variables (#99367)

avikchaudhuri committed 1 year ago

c75ac11f

[Profiler][Easy] Fix typo in Profiler report input shapes (#99430)

aaronenyeshi committed 1 year ago

237f917f

fix osd rank0_only in fsdp (#99136)

ZhongYingMatrix committed 1 year ago

af7fed1d

[memory allocator] fix ifdef typo (#99553)

zdevito committed 1 year ago

2402fe52

Add device_asserts before indirect loads and stores (#98590)

lezcano committed 1 year ago

495e1b4d

Make opacus xfail instead of skip (#99380)

wconstab committed 1 year ago

9ac2b041

Revert "Use correct standard when compiling NVCC on Windows (#99492)"

pytorchmergebot committed 1 year ago

cfacb5ea

[SPMD][Easy] switch to tree_map_only to simplify code (#99547)

mrshenli committed 1 year ago

ca89e794

Use correct standard when compiling NVCC on Windows (#99492)

r-barnes committed 1 year ago

db694456

[c10d] Faster coalescing (#98793)

kwen2501 committed 1 year ago

db456ab8

Run post-aot compiler at compilation time, not at runtime. (#99457)

ezyang committed 1 year ago

bc9eaa7a

[BE] Refactoring test execution and improving comments (#99467)

ZainRizvi committed 1 year ago

75469725

[Composable API] Add fully_shard debug function to print sharded tree structure, module names and managed param fqns (#99133)

zhaojuanmao committed 1 year ago

6ca991ca

Warn if guards are added to ShapeEnv after we produced guards (#97820)

ezyang committed 1 year ago

6b6dc441

Fix Tensor.uniform_ documentation to mention generator argument (#99510)

tfsingh committed 1 year ago

2aa35e6c

[fx] Variatic arg matching (#99431)

angelayi committed 1 year ago

d6d55f85

improve mkldnn matmul performance when one input is contiguous tensor but the strides is not default contiguous strides (#99511)

XiaobingSuper committed 1 year ago

e21f648c

make ATen/native/cuda/UpSampleNearest3d.cu data_ptr-correct (#99328)

mikey dagitses committed 1 year ago

efa16c20

Newer Older