pytorch/xla

Pull Requests Commits

Reenable the distributed checkpointing test (#8424)

JackCaoG committed 297 days ago

Verified 591c3979

Support SegmentID when doing data prallel SPMD (#8425)

JackCaoG committed 301 days ago

Verified 1c91219f

Use regular `torch.Tensor` for CPU tensors (#8416)

qihqi committed 301 days ago

Verified 20f51661

Update XLA pin to 11/22 (#8411)

JackCaoG committed 303 days ago

Verified 39e67b5d

scan and scan_layers (#7901)

tengyifei committed 306 days ago

Verified 51575db4

Add an example that train the torchtitan version of llama. (#8400)

qihqi committed 306 days ago

Verified 31d348e5

Force specialize_float = True for torch_xla (#8404)

ezyang committed 307 days ago

Verified d503ca54

Only write to HBM at the last iteration. (#8393)

vanbasten23 committed 308 days ago

Verified d572aeb5

Unify how cpp test is being run (#8391)

JackCaoG committed 309 days ago

Verified 2ec22641

Update ARC runner version to 2.321.0 (#8397)

mbzomowski committed 309 days ago

Verified fbed916a

Update XLA pin to 11/17 (#8392)

JackCaoG committed 310 days ago

Verified 0c3d54bf

Fix the fallback op in SPMD (#8386)

JackCaoG committed 310 days ago

Verified defd0a9d

[torch_xla2] add `nn.functional.multilabel_margin_loss` ... `randint` (#8378)

ManfeiBai committed 313 days ago

Verified 91f5c8ad

Fix incorrect xla type for view_as_complex (#8382)

JackCaoG committed 313 days ago

Verified bd4006ef

update the path to tsl/protobuf/error_codes.pb.h (#8383)

JackCaoG committed 313 days ago

Verified 5b3c3315

Support the case where query_len>context_len in the multi-queries paged attention. (#8356)

vanbasten23 committed 315 days ago

Verified 102cd489

Expose lazy all number special scalars (#8361)

rpsilva-aws committed 315 days ago

Verified f610f704

[torch_xla2] Enable nn.functional.cosine_embedding_loss (#8368)

ManfeiBai committed 315 days ago

Verified d4c90c91

Fix op info test for svd_lowrank (#8376)

dvhg committed 315 days ago

Verified 7220aee4

Revert S64 and U64 type change for Neuron device due to hardcoded dtype checks in torch-xla (#8375)

aws-nm9 committed 315 days ago

Verified bc227f7f

Remove unnecessary prints in PagedAttention (#8374)

WoosukKwon committed 316 days ago

Verified baa14f36

Fix XLA type of the view_as_real (#8370)

JackCaoG committed 316 days ago

Verified 071ddfa6

Improve some typing annotations (#8369)

tengyifei committed 317 days ago

Verified c4f27718

Add tengyifei to docs code owners (#8365)

tengyifei committed 321 days ago

Verified 5a4641b4

Add mikegre-google as docs owner (#8364)

tengyifei committed 321 days ago

Verified 09f36bee

XLA downcast for S64 and U64 for Neuron (#8347)

aws-nm9 committed 321 days ago

Verified 03f07e2a

Add aten::lift_fresh to the exception list that converts input to torch tensor (#8360)

barney-s committed 322 days ago

Verified 6454b42f

Add inplace ops for sigmoid and tanh (#8358)

barney-s committed 322 days ago

Verified 77abab25

Revive "Modularize zero step function and make it customizable" #7233 (#8332)

jeffhataws committed 323 days ago

Verified 7d6a4f25

Make XLA tests export new IR compatible (#8354)

tugsbayasgalan committed 323 days ago

Verified 81c4caa7

Newer Older