pytorch/xla

Pull Requests Commits

change back file importer

yaochengji committed 271 days ago

451db0c4

more test for aten.linear

yaochengji committed 271 days ago

10c124df

Fix aten.linear and add aten.einsum

yaochengji committed 272 days ago

13276026

Dynamo/AOTAutograd traceable flash attention (#8654)

zpcore committed 274 days ago

Verified 9ae017e2

openxla pin update to 20250131 (#8621)

zpcore committed 274 days ago

Verified 82d35047

[torchax] fix einsum op conversion (#8657)

yaochengji committed 274 days ago

Verified 8572e750

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

rpsilva-aws committed 275 days ago

Verified 39dd7959

Add a stub for host offloading docs (#8656)

tengyifei committed 275 days ago

Verified e583c2cc

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

tengyifei committed 276 days ago

Verified 93a2ba6b

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

tengyifei committed 276 days ago

Verified d9593f03

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

yaochengji committed 277 days ago

Verified 7243b808

Build v2.6 release wheel (#8645)

tengyifei committed 277 days ago

Verified 01839974

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

qihqi committed 277 days ago

Verified 25d09fdd

Prepare v2.6.0-rc10 (#8639)

tengyifei committed 277 days ago

Verified 310d2d81

scan documentation (#8631)

tengyifei committed 278 days ago

Verified 225c65bd

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

tengyifei committed 278 days ago

Verified 8e6ca600

Rename torch_xla2 to torchax (#8599)

qihqi committed 280 days ago

Verified 8b241409

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

vealocia committed 281 days ago

Verified 79e4e72f

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

bhavya01 committed 281 days ago

Verified aef5f6ea

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

jeffhataws committed 281 days ago

Verified e694c90d

Prepare to cut r2.6 rc9 (#8619)

tengyifei committed 282 days ago

Verified 6f367dfa

Improve build_developer.sh (#8579)

tengyifei committed 283 days ago

Verified d382feed

use env to skip PJRT initialize (#8609)

zpcore committed 283 days ago

Verified 557d9f34

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

fenghuizhang committed 283 days ago

Verified 9a1405d6

Introducing experimental gradient accumulation API (#8584)

rpsilva-aws committed 283 days ago

Verified 36dcba32

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

fenghuizhang committed 283 days ago

Verified 5b877beb

Add two placeholder example directories (#8601)

tengyifei committed 283 days ago

Verified fbbdfca1

training UX: automatic generating make_train_step (#8495)

qihqi committed 284 days ago

Verified f5b33c5d

Prepare to create 2.6.0rc8 builds (#8598)

tengyifei committed 284 days ago

Verified bbd66b93

Pipes attn_logits_soft_cap through multi_queries_paged_attention (#8583)

fenghuizhang committed 284 days ago

Verified 6b785fc8

Older