pytorch/xla

Pull Requests Commits

Switch to GetOperandListWithToken naming for func GetOperandList

jeffhataws committed 1 year ago

8dc7af94

jeffhataws committed 1 year ago

fae4d071

jeffhataws committed 1 year ago

6fd0d0a7

Use token_handler.GetInput to capture token

jeffhataws committed 1 year ago

49f4ba4e

Add separate BuildAllGatherCoalesced builder and AllGatherCoalesced class

jeffhataws committed 1 year ago

3cd39ea5

jeffhataws committed 1 year ago

8e6715e7

Separate out the reduce-scatter-coalesce changes into a separate PR

jeffhataws committed 1 year ago

5843b434

Fix reduce-scatter-coalesce to be compatible with openxla reduce-scatter tuple change without token

jeffhataws committed 1 year ago

f9f3a712

Fix allgather to be compatible with openxla allgather tuple change without token

jeffhataws committed 1 year ago

ec27f90d

Code/test fixes to get run_tests.sh to run on CPU

jeffhataws committed 1 year ago

3e97aa84

Fix "SyntaxError: 'return' outside function" error

jeffhataws committed 1 year ago

cb5b6fbe

clang-format-7 and python lint fixes

jeffhataws committed 1 year ago

63431828

Add all-gather and reduce-scatter coalescence support for FSDP.

jeffhataws committed 1 year ago

57e6b366

[SPMD] fix bug with XLAShardedTensor.__repr__ (#5807)

yeounoh committed 1 year ago

Verified 6addbde3

Use upstream XLA concurrency utilities (#5799)

will-cromar committed 1 year ago

Verified 6e661304

Add python 3.11 trigger for 2.1 release build (#5810)

JackCaoG committed 1 year ago

Verified a3bbbb1e

Enable passing down dynamic dimensions from torch to XLA (#5790)

lsy323 committed 1 year ago

Verified a0a048e1

Don't set $TPU_LIBRARY_PATH during import

will-cromar committed 1 year ago

Verified d8c6a83f

Record the lazy tracing time(C++) in metrics (#5757)

JackCaoG committed 1 year ago

Verified 58ce7e17

Use TPU profiler plugin (#5793)

will-cromar committed 1 year ago

Verified 3b34cb27

Add GKE support and various usability improvements in CheckpointManager (#5770)

jonb377 committed 1 year ago

Verified 5710a83c

Lower AtenFull op (#5781)

danielvegamyhre committed 1 year ago

Verified b3ed82c9

Transfer data directly to the device (#5772)

will-cromar committed 1 year ago

Verified 05a3cdda

[SPMD] suppor DTensor API integration (#5776)

yeounoh committed 1 year ago

Verified e60428d1

disable test_set temporarily (#5792)

yeounoh committed 1 year ago

Verified 6f4183c8

Make the pjrt gpu allocator configurable (#5759)

anw90 committed 1 year ago

Verified 56733fbc

Remove some unused code from `csrc/runtime` (#5785)

will-cromar committed 1 year ago

Verified 79557ccd

Implement mark_sharding as a custom op to support dynamo spmd activation sharding (#5712)

wonjoo-wj committed 1 year ago

Verified 367f47f5

Use default method in stablehlo txt/bytecode getter of StableHLOGraphModule (#5745)

lsy323 committed 1 year ago

Verified d2f389e7

Dsiable C++ test correctly (#5783)

JackCaoG committed 1 year ago

Verified d5d02306

Older