strisunshinewentingwang/pytorch

Pull Requests Commits

update to latest kineto

briancoutinho committed 2 years ago

27e5b868

add kineto logs for debugging

briancoutinho committed 2 years ago

304ccc22

briancoutinho committed 2 years ago

25c65050

briancoutinho committed 2 years ago

a27d46ea

Brian Coutinho committed 2 years ago

9ddc229c

point to latest kineto -https://github.com/pytorch/kineto/pull/724

briancoutinho committed 2 years ago

dc296799

minor change to improve print order

briancoutinho committed 2 years ago

e2279296

[kineto] Enable CUPTI range performance metrics profiling in pytorch build

briancoutinho committed 2 years ago

08b7c8e6

[dtensor] remove unused tests and fix ci (#97064)

wanchaol committed 2 years ago

4805441b

Revert "[inductor] do benchmark in sub processes for max autotuning (#96410)" (#97075)

huydhn committed 2 years ago

a5923ab3

component-level configurable logging for dynamo, inductor, aot (#94858)

mlazos committed 2 years ago

a1c46e5f

Add new parameter `materialize_grads` to torch.autograd.grad() (#97015)

zqigolden committed 2 years ago

086ce765

[inductor] do benchmark in sub processes for max autotuning (#96410)

shunting314 committed 2 years ago

34256bc7

Update MHA doc string (#97046)

Michael Gschwind committed 2 years ago

b1322203

[Inductor] Eliminate redundant to_dtype node (#96650)

EikanWang committed 2 years ago

915cbf82

Use is_available instead of device_count to check for CUDA availability (#97043)

huydhn committed 2 years ago

679dec84

Increase the timeout value for linter calculate-docker-image (#96993)

huydhn committed 2 years ago

c62fc81c

[docs] passing LogSoftmax into NLLLoss (#97001)

jamesbraza committed 2 years ago

b390e703

Remove obsolete "merge -g" flag from update_commit_hashes.py (#97033)

kit1980 committed 2 years ago

410210b3

Re-enable test_ops_jit on Windows (#96859) (#96931)

huydhn committed 2 years ago

db2c1ea8

[dynamo][dashboard] fix triton clone step in dashboard (#96623)

davidberard98 committed 2 years ago

a4c706bc

Make keep-going work for more than linux (#96974)

janeyx99 committed 2 years ago

4a90aca6

Fix CPU bitwise shifts for out-of-limit shift values (#96659)

bjhargrave committed 2 years ago

b59a60dd

Remove unnecessary items() call in zero_grad (#97040)

Skylion007 committed 2 years ago

dd9ade63

[SDPA] Remove the chunk_grad from mem-eff attention (#96880)

drisspg committed 2 years ago

98a5cf09

Fail fast when dynamo attempts to add unspecialized int/float as additional graph inputs (#96786)

guangy10 committed 2 years ago

d4b8ed2b

Improve size mismatch error messaging referencing mat/vet sizes (#96863)

kiersten-stokes committed 2 years ago

cea13ad9

Bind increment_version to python (#96852)

albanD committed 2 years ago

985fc66b

Fixed print tensor.type() issue. (#96381)

NmomoN committed 2 years ago

1983b317

[static-runtime] one more attempt to improve crash log readability (#96903)

tenpercent committed 2 years ago

57bb5b15

Older