strisunshinewentingwang/pytorch

Pull Requests Commits

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

Revert "[cuda rng] Making offset calculation independent of device properties (#98988)"

pytorchmergebot committed 1 year ago

5cb788a9

Skip vision_maskrcnn after #98923 (#99394)

huydhn committed 1 year ago

5d395769

Fix (4 device) multi-gpu `ShardedGradScaler` Tests in `ciflow/periodic` (#99485)

speediedan committed 1 year ago

6b8bab8e

[reland][quant][fix] Compare resnet with quantizer api with the prepare_fx and decomposed convert flow (#99355)

jerryzh168 committed 1 year ago

b0df0cd7

make ATen/native/cuda/TensorModeKernel.cu data_ptr-correct (#99330)

mikey dagitses committed 1 year ago

391a3add

make ATen/native/cuda/ReflectionPad.cu data_ptr-correct (#99325)

mikey dagitses committed 1 year ago

8eb77434

make ATen/native/cuda/Col2Im.cu data_ptr-correct (#99348)

mikey dagitses committed 1 year ago

4d3011b6

make ATen/native/cuda/Shape.cu data_ptr-correct (#99343)

mikey dagitses committed 1 year ago

121edd21

Add typing to debug_utils and repro (#99452)

ezyang committed 1 year ago

b01edf45

Refactor debug_utils into after_aot and after_dynamo modules (#99450)

ezyang committed 1 year ago

2e25fb5d

Codegen fixed size for static sympy values (#99362)

voznesenskym committed 1 year ago

a3ee9800

[SPMD] Add sym_stride to DSymInt (#99504)

mrshenli committed 1 year ago

e605b5df

[SPMD] Support DSymInt for slice_backward in SPMD expansion (#99501)

mrshenli committed 1 year ago

2cb8a8d4

[SPMD] Support SymInt with non-op call_function nodes (#99420)

mrshenli committed 1 year ago

29229614

Newer Older