strisunshinewentingwang/pytorch

Pull Requests Commits

[Dont review] default partitioner for recomputable ops

anijain2305 committed 1 year ago

827f61ff

[min-cut partitioner] Disable a heuristic if graph has recomputable ops

anijain2305 committed 1 year ago

e58aeb8c

Delete ifdyn and ifunspec combinators (#103596)

ezyang committed 1 year ago

ddf4cd69

Add `generator` argument in `torch.randn` signature (#102075)

liang-hou committed 1 year ago

e82616d9

Link torch.cat in docstring of torch.stack and vice versa (#103421)

moi90 committed 1 year ago

a0885dff

Support params/buffers inside cond and map (#102310)

tugsbayasgalan committed 1 year ago

766f236b

add instruction to compile with new C++ ABI (#95177)

jingxu10 committed 1 year ago

600f7dc2

Improve DDPOptimizer Logging (#103489)

wconstab committed 1 year ago

55cf5c00

Silence `has_cuda` deprecation in optim (#103610)

awgu committed 1 year ago

9152d0e5

[Pytorch] aten::stack (#103344)

lucylq committed 1 year ago

d0ff640e

Fix composable `checkpoint(use_reentrant=True)` with multi args (#103590)

awgu committed 1 year ago

2eea3cb1

[inductor] Fix an expression printer issue during generate_return (#103557)

desertfire committed 1 year ago

c2952e8b

Update optimizer tests to compile with fullgraph (#103559)

mlazos committed 1 year ago

dc3fa9e5

[FSDP][4/n]Update use_dtensor option for _optim_utils.py (#103599)

wz337 committed 1 year ago

7dd0f525

[activation checkpoint][dynamo] Wrap AC into Tag based higher order op (#102935)

anijain2305 committed 1 year ago

bd0ed940

[activation checkpointing] Tagging based min cut partitioner (#103357)

anijain2305 committed 1 year ago

df050574

Move locals/globals to output graph, make it easier to access them anywhere (#103456)

voznesenskym committed 1 year ago

aece6705

Simple Source traversal util (#103450)

voznesenskym committed 1 year ago

d27bc34f

Update clang-tidy install in CONTRIBUTING.md (#101247)

arunppsg committed 1 year ago

6db21a9c

Continue simplifying dynamic shapes tests (#103592)

ezyang committed 1 year ago

99464992

[PT2][Quant] Change quat conv bn fusion code (#103556)

kimishpatel committed 1 year ago

49dcf48e

Revert "Add groups to dynamo benchmarking output data (#103268)"

pytorchmergebot committed 1 year ago

a60f6dbe

optimize reflection padding performance on CPU (#102254)

mingfeima committed 1 year ago

69b09eca

[inductor] use aten.kernel.OVERLOAD_NAME instead of aten.kernel in python wrapper (#103576)

davidberard98 committed 1 year ago

717e63b7

[Dynamo] VariableTracker.recursively_contains should be updated correctly when mutation happens (#103564)

yanboliang committed 1 year ago

5c3556da

[_memory_viz.py] Fix bug when using profile_plot (#103384)

zdevito committed 1 year ago

0ca3c6f7

Revert "Register more foreach op lowerings (#102654)"

pytorchmergebot committed 1 year ago

6ff6b490

[inductor] Fix no-xdim reductions (#103527)

peterbell10 committed 1 year ago

b1adaa87

[DDP] multiple forward support for static graph (#103487)

rohan-varma committed 1 year ago

80139fc2

[DDP] Refactor _DDPSink to take DDP weakref (#103304)

rohan-varma committed 1 year ago

780b24b2

Older