pytorch/xla

Pull Requests Commits

fix a bug in tensor.flatten

yaochengji committed 273 days ago

4e5c987e

Install `clang-17` on development Dockerfile. (#8673)

ysiraichi committed 273 days ago

Verified ec10537c

[CI] Fix torchbench setup failure (#8670)

lsy323 committed 274 days ago

Verified a4356250

Add ragged paged attention (#8659)

vanbasten23 committed 274 days ago

Verified 84800944

`cummax`: fix 0-sized dimension reduction. (#8653)

ysiraichi committed 274 days ago

Verified 1c537904

[torchax][distributed] Use local device for default jax device (#8667)

lsy323 committed 274 days ago

Verified 35789406

Dynamo/AOTAutograd traceable flash attention (#8654)

zpcore committed 277 days ago

Verified 9ae017e2

openxla pin update to 20250131 (#8621)

zpcore committed 277 days ago

Verified 82d35047

[torchax] fix einsum op conversion (#8657)

yaochengji committed 277 days ago

Verified 8572e750

Fix the retention for existing gradients in the Grad Acc API (#8658)

rpsilva-aws committed 278 days ago

Verified 39dd7959

Add a stub for host offloading docs (#8656)

tengyifei committed 278 days ago

Verified e583c2cc

Fix C++11 ABI wheel instructions (#8650)

tengyifei committed 279 days ago

Verified 93a2ba6b

Update README.md for PyTorch/XLA 2.6 release (#8647)

tengyifei committed 279 days ago

Verified d9593f03

Fix metrics saver when a file location is specified (#8644)

yaochengji committed 280 days ago

Verified 7243b808

Build v2.6 release wheel (#8645)

tengyifei committed 280 days ago

Verified 01839974

Add a lowering to linear instead of using torch decomp (#8643)

qihqi committed 280 days ago

Verified 25d09fdd

Prepare v2.6.0-rc10 (#8639)

tengyifei committed 280 days ago

Verified 310d2d81

scan documentation (#8631)

tengyifei committed 281 days ago

Verified 225c65bd

Fix grad accumulation tests on TPUv5/v6 (#8628)

tengyifei committed 281 days ago

Verified 8e6ca600

Rename torch_xla2 to torchax (#8599)

qihqi committed 283 days ago

Verified 8b241409

Make partition_spec as tuple to fix issue #8522 (#8613)

vealocia committed 284 days ago

Verified 79e4e72f

Update README.md to recommend torch_xla cxx11 builds by default. (#8624)

bhavya01 committed 284 days ago

Verified aef5f6ea

Improve the nprocs value error message to be actionable (#8622)

jeffhataws committed 284 days ago

Verified e694c90d

Prepare to cut r2.6 rc9 (#8619)

tengyifei committed 285 days ago

Verified 6f367dfa

Improve build_developer.sh (#8579)

tengyifei committed 286 days ago

Verified d382feed

use env to skip PJRT initialize (#8609)

zpcore committed 286 days ago

Verified 557d9f34

Fix the signature of paged_attention by marking megacore_mode optional (#8615)

fenghuizhang committed 286 days ago

Verified 9a1405d6

Introducing experimental gradient accumulation API (#8584)

rpsilva-aws committed 286 days ago

Verified 36dcba32

Fix an issue when piping attn_logits_soft_cap through in vllm. (#8600)

fenghuizhang committed 286 days ago

Verified 5b877beb

Add two placeholder example directories (#8601)

tengyifei committed 286 days ago

Verified fbbdfca1

Older