pytorch/xla

Pull Requests Commits

Update troubleshooting doc with few common env var combination

JackCaoG committed 1 year ago

16bed80e

[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

Refactor type conversion functions out of `//torch_xla/csrc:tensor` (#5777)

will-cromar committed 1 year ago

Verified ee210bb7

Disable flaky cpp test (#5779)

JackCaoG committed 1 year ago

Verified 963204a3

fixing neuron import (#5775)

aws-kingrj committed 1 year ago

Verified bd29e796

Support autocheckpointing (#5753)

jonb377 committed 1 year ago

Verified 46643801

fix squeeze op lowering issue when dim is not in sorted order (#5751)

zpcore committed 1 year ago

Verified 21920318

Include Dynamo executation in the executation cause analysis (#5758)

JackCaoG committed 1 year ago

Verified b8b8ea86

Remove `_unsafe_index` implementation. (#5769)

ysiraichi committed 1 year ago

Verified 28c277f0

update doc to use PJRT_DEVICE=CUDA instead of PJRT_DEVICE=GPU (#5754)

vanbasten23 committed 1 year ago

Verified f01cdb60

Revert "Transfer data directly to the device (#5752)" (#5765)

will-cromar committed 1 year ago

Verified a93e14bf

Transfer data directly to the device (#5752)

will-cromar committed 1 year ago

Verified 5ca36cf5

Support PreemptionSyncManager in XlaCoordinator (#5733)

jonb377 committed 1 year ago

Verified b20a082a

Correct the multinode training doc (#5747)

vanbasten23 committed 1 year ago

Verified 4038f8e7

Older