pytorch/xla

Pull Requests Commits

Fix distributed checkpoint errors when PyTorch CUDA is enabled. (#6421)

ysiraichi committed 1 year ago

Verified 85c66f5f

Revert "aggregate.py: add tab format" (#6424)

cota committed 1 year ago

Verified 010b6f09

aggregate.py: add tab format (#6405)

cota committed 1 year ago

Verified 90d9dd1d

[benchmarks] Fix CUDA device check. (#6419)

ysiraichi committed 1 year ago

Verified be44bc6b

Use PyTorch's dynamo benchmark skip-list. (#6416)

ysiraichi committed 1 year ago

Verified 492fe27e

Fix some more core aten ops (#6414)

wonjoo-wj committed 1 year ago

Verified a83be44c

[torchbench] Re-land: Force `llama_v2_7b_16h` to use SGD. (#6402)

ysiraichi committed 1 year ago

Verified 0baf1dae

Move torch import closer to the place it uses (#6412)

alanwaketan committed 1 year ago

Verified ba6a9f5b

nightly.sh: pass --always to git describe --tags (#6406)

cota committed 1 year ago

Verified 9334ebbe

nightly.sh: add flag to set Hugging Face token (#6410)

cota committed 1 year ago

Verified 4fc697cb

Revert: #6368 to fix nightly CI coverage run. (#6369)

yeounoh committed 1 year ago

Verified 48a1ae3d

Point to v2.2.0 release (#6418)

zpcore committed 1 year ago

Verified f66406bb

Update OpenXLA pin to 20240124 (#6390)

vanbasten23 committed 1 year ago

Verified 0aee211d

[Pallas] PoC Integration (#6340)

alanwaketan committed 1 year ago

Verified 56db8f23

[FSDPv2] Add a user guide (#6386)

alanwaketan committed 1 year ago

Verified de5d7647

aggregate.py: allow picking backends to report on (#6392)

cota committed 1 year ago

Verified 2b7f4354

Force to fp16 models to fp32 if `XLA_USE_FP16` is already set. (#6389)

ysiraichi committed 1 year ago

Verified 2a17ac20

Define PJRT plugin interface in C++ (#6360)

will-cromar committed 1 year ago

Verified bd95eb13

Add compilation time analysis. (#6371)

golechwierowicz committed 1 year ago

Verified 32d24adb

Revert "Lower RandPerm" (#6394)

JackCaoG committed 1 year ago

Verified 2f4275f5

benchmarks/nightly.sh: add -P flag to enable profiling (#6298)

cota committed 1 year ago

Verified 259b7dcd

aggregate.py: add --baseline option (#6349)

cota committed 1 year ago

Verified 77b9b2da

Reapply "Fix some more core aten ops (#6342)" (#6377) (#6387)

cota committed 1 year ago

Verified 9e4db963

Update DDP example in README.md (#6384)

bhavya01 committed 1 year ago

Verified c70e4cc6

Lower RandPerm (#6315)

wonjoo-wj committed 1 year ago

Verified e8b1de97

Add GPU plugin build to CI (#6373)

will-cromar committed 1 year ago

Verified e72cf490

Support PJRT C API `create_options` (#6289)

will-cromar committed 1 year ago

Verified c738156d

Provide op lowering for Logit (#6357)

bhavya01 committed 1 year ago

Verified a4d2d764

Use `benchmark_cls` for checking precision.` (#6375)

ysiraichi committed 1 year ago

Verified a1e51e4c

Fix walrus operator precedence. (#6374)

ysiraichi committed 1 year ago

Verified 44660d8b

Newer Older