pytorch/xla

Pull Requests Commits

Include both pytorch and torch_xla revisions in the compilation cache key. (#9383)

zhanyong-wan committed 1 year ago

Verified aaff9591

Add autocast feature as `torchax.amp.autocast`. (#9364)

qihqi committed 1 year ago

Verified 89cc066a

Update the tuned block size. (#9376)

QiliangCui committed 1 year ago

Verified f20b89f9

Deprecate ShapeOfXlaOp in favor of GetShape (#9381)

zhanyong-wan committed 1 year ago

Verified b737b4a1

Fix race condition in runAtenTest (#9306)

benawilson committed 1 year ago

Verified f8bd6c48

Improve the API for binding python functions in C++ (#9351)

zhanyong-wan committed 1 year ago

Verified 86c59725

update l_ref in kernel matrix calculation (#9372)

Chenyaaang committed 1 year ago

Verified 77c40f3f

Avoid recompilation caused by `scan_layers` (#9367)

tengyifei committed 1 year ago

Verified 41b9e452

fix all_gather_into_tensor test and logic (#9332)

bfolie committed 1 year ago

Verified 9b0b02f0

Improve `assume_pure` SPMD functionality (#9360)

tengyifei committed 1 year ago

Verified 0be0fbdf

Add an option to not use dlpack. (#9304)

qihqi committed 1 year ago

Verified 55a75404

docs: fix contribute guide command (#9355)

giacomoni committed 1 year ago

Verified 20899c72

Update ragged paged attention kernel to prevent vmem oom (#9346)

Chenyaaang committed 1 year ago

Verified bea86eec

Split _ops and _decomps. (#9323)

qihqi committed 1 year ago

Verified 3a1ed628

Ignore the MODULE.bazel* files generated by bazel. (#9352)

zhanyong-wan committed 1 year ago

Verified 15366e94

Switch XLA_CHECKs to ABSL_CHECKs in lowering_context. (#9338)

zhanyong-wan committed 1 year ago

Verified c2aa21ee

cleanup: remove defunct --distinct_host_configuration flag (#9300)

rickeylev committed 1 year ago

Verified 1d409399

Fix usage of dimensions_size() to check for tuple (#9347)

bfolie committed 1 year ago

Verified 49b0b3b3

Surface CUDA deprecation warning. (#9333)

ysiraichi committed 1 year ago

Verified edd38caa

Update pins (#9330)

pgmoka committed 1 year ago

Verified 503f8280

doc: update pytorch-on-xla-devices and troubleshoot doc for tensor synchronization issue (#9258)

aws-yyjau committed 1 year ago

Verified 3bc9ece7

Update submodules whenever we switch branches. (#9320)

zhanyong-wan committed 1 year ago

Verified 1bd3042f

Remove nightly CUDA builds (#9329)

tengyifei committed 1 year ago

Verified ed14590a

Migrate `.to(torch_xla.device())` to `.to('xla')` (#9324)

ghpvnist committed 1 year ago

Verified 78cff036

Migrate uses of `import torch_xla as xla` to `import torch_xla` (#9325)

ghpvnist committed 1 year ago

Verified ae4f06b3

Revert Create copy of nightly wheel without version number (#9318)

bfolie committed 1 year ago

Verified f921e5e8

Clean up torchax readme. (#9321)

zhanyong-wan committed 1 year ago

Verified e38d2807

enable_python_dispatcher in some XLA custom passes. (#9312)

laithsakka committed 1 year ago

Verified 9a517a95

Test torchprime from PyTorch/XLA (#9152)

tengyifei committed 1 year ago

Verified 728043e6

Extend build_developer.sh and git_sync_main.py with a -a flag. (#9319)

zhanyong-wan committed 1 year ago

Verified 37ed591d

Newer Older