strisunshinewentingwang/pytorch

Pull Requests Commits

Add runtime assertions for input shape constraints (#100247)

tugsbayasgalan committed 1 year ago

9b3552eb

[inductor] Enable conditional use of tl.reduce (#100569)

peterbell10 committed 1 year ago

8f1122ce

[inductor] Stop using `x + tl.zeros(...)` in generated triton (#100163)

peterbell10 committed 1 year ago

94d306fd

[inductor] Fix argmin/max with duplicate values (#100573)

peterbell10 committed 1 year ago

06fbd5dc

[inductor] Fix nan-handling of max and min reductions (#100572)

peterbell10 committed 1 year ago

49189401

Fix a number of issues with divs in ValueRangeAnalysis (#100547)

lezcano committed 1 year ago

19a57870

[c10d] Fix subprocess group handlig in scatter_object_list. (#100552)

Rodrigo Kumpera committed 1 year ago

a204f7f5

[vmap] bucketize (#95783)

kshitij12345 committed 1 year ago

aecbaa5d

Revert "[export] Pickle result of export (#100423)"

pytorchmergebot committed 1 year ago

c4fd76e7

[export] Pickle result of export (#100423)

angelayi committed 1 year ago

7226dbcb

[exportdb] Change case ids to case names for UserErrors. (#100600)

zhxchen17 committed 1 year ago

8d598f2f

Don't build Triton from source in benchmarks/dynamo/Makefile (#100613)

ezyang committed 1 year ago

c58d9642

[docs] Docs for writing ATen IR passes + FX Pattern matching (#100577)

angelayi committed 1 year ago

8eb82135

Add AotAutogradFallbackTests to dynamic suite (#100454)

voznesenskym committed 1 year ago

fe3ecfe0

Reland basic dynamo support for traceable collectives (#100476)

wconstab committed 1 year ago

2dca4181

Fix graph break in a common `func(self, *args)` pattern (Faster stable diffusion) (#100444)

voznesenskym committed 1 year ago

9f3c6b1b

Improve minifier printing to be more chatty when it makes sense (#100486)

ezyang committed 1 year ago

c2556c03

Misc accuracy improvements on minifier (#100447)

ezyang committed 1 year ago

c7e9f406

[codemod] Replace hasattr with getattr in caffe2/test/distributed/fsdp/test_fsdp_optim_state.py (#100360)

r-barnes committed 1 year ago

7f997aa3

[vision hash update] update the pinned vision hash (#100510)

pytorchmergebot committed 1 year ago

8df748f3

Fix bug in process_group_name when there is duplicate pgs (#100518)

xw285cornell committed 1 year ago

c29ab841

[replicate] input casting support (#100216)

rohan-varma committed 1 year ago

253b9d32

[primTorch] add ref for `polar` (#100345)

nkaretnikov committed 1 year ago

e87ed2a8

Ensure device arg is passed to test_transformers (#100260)

drisspg committed 1 year ago

2892c06e

inplace PyTorchStreamReader getRecord() (#100418)

jiayisuse committed 1 year ago

f558bb6f

Use Boxed Calling Convention for AOT Eager (#100417)

eellison committed 1 year ago

e6c0164f

Require contiguous for view_as_complex (#100428)

eellison committed 1 year ago

d67e4db8

Remove speech_transformer workaround, torchbench handles it correctly now (#100558)

ezyang committed 1 year ago

d25c93f9

[dynamo] Minor fixes and clean-up in eval_frame.c (#100496)

colesbury committed 1 year ago

fd841763

add checkpoint support for custom device (#99626)

heidongxianhua committed 1 year ago

6aeb85ad

Newer Older