pytorch/xla

Pull Requests Commits

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

fenghuizhang committed 284 days ago

Verified 9a1405d6

Introducing experimental gradient accumulation API (#8584)

rpsilva-aws committed 284 days ago

Verified 36dcba32

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

fenghuizhang committed 284 days ago

Verified 5b877beb

Add two placeholder example directories (#8601)

tengyifei committed 284 days ago

Verified fbbdfca1

training UX: automatic generating make_train_step (#8495)

qihqi committed 285 days ago

Verified f5b33c5d

Prepare to create 2.6.0rc8 builds (#8598)

tengyifei committed 285 days ago

Verified bbd66b93

Pipes attn_logits_soft_cap through multi_queries_paged_attention (#8583)

fenghuizhang committed 285 days ago

Verified 6b785fc8

prepare rc7 for r2.6 with libtpu update (#8586)

ManfeiBai committed 289 days ago

Verified a295f7d0

Re-introduce "XLA_USE_32BIT_LONG" flag (#8571)

rpsilva-aws committed 291 days ago

Verified 78fa7576

Prepare to make rc6 wheel (#8582)

tengyifei committed 291 days ago

Verified 18efb6dc

[scan] Reduce memory usage (#8562)

tengyifei committed 291 days ago

Verified dc768796

Update terraform providers for tpu-pytorch (#8575)

tengyifei committed 291 days ago

Verified 02602094

Improve and refine MLP tests for extensibility and A/B testing (#8561)

rpsilva-aws committed 291 days ago

Verified 1d615566

Fix mark_sharding logic (#8578)

yaochengji committed 291 days ago

Verified 14acab7a

Correct libtpu_storage_path for stable libtpu builds (#8576)

tengyifei committed 292 days ago

Verified 79de480f

Prepare to make rc5 wheel (#8573)

tengyifei committed 292 days ago

Verified e513e132

Update iam.auto.tfvars (#8570)

tengyifei committed 292 days ago

Verified 4edbf61d

pin update (#8559)

lsy323 committed 293 days ago

Verified 1c89675e

Set gradient_as_bucket_view=True in test_train_mp_imagenet (#8558)

tengyifei committed 293 days ago

Verified 2fdf721f

fix batch_norm amp autocast (#8556)

yaochengji committed 293 days ago

Verified fda4ecae

[scan] Test we don't recompile under debugging env flags (#8555)

tengyifei committed 293 days ago

Verified 79a7de0c

Introduce HLO graph bindings (#8551)

rpsilva-aws committed 296 days ago

Verified 3784b268

Always run TPU tests on new PRs (#8552)

tengyifei committed 296 days ago

Verified 6963e195

Metadata agnostic user computation hash (#8550)

rpsilva-aws committed 296 days ago

Verified 6e5659ec

Lower cummax op (#8491)

zyy-martin committed 296 days ago

Verified c394d1b5

[torch_xla2] Fix reenabled op info tests (#8548)

dvhg committed 296 days ago

Verified af223d32

[Computation Hash] Introduce deterministic hash for user computations (#8539)

rpsilva-aws committed 297 days ago

Verified 196cab3f

Revert "fix batch_norm amp autocast" (#8547)

tengyifei committed 297 days ago

Verified 05ca2241

Fix typos in pytorch on xla docs (#8543)

MichalPitr committed 298 days ago

Verified de756865

fix batch_norm amp autocast (#8498)

yaochengji committed 298 days ago

Verified 1cdb1eff

Newer Older