pytorch/xla

Pull Requests Commits

Update libtpu pin to 2024/12/05 / 0.0.5 stable (#8461)

tengyifei committed 1 year ago

Verified 85cf18ac

Skip the computation if the current block lies above the causal mask diagonal. (#8448)

vanbasten23 committed 1 year ago

Verified 5270d7c4

Stricter testing of sharding annotations after scan (#8456)

tengyifei committed 1 year ago

Verified a48ff6a1

DLPack: fix test using PyTorch API. (#8348)

ysiraichi committed 1 year ago

Verified 4c99d21c

Fix an edge case issue when storing output in the paged attention. (#8431)

vanbasten23 committed 1 year ago

Verified a1a11457

Update torch version to 2.5.1 (#8452)

qihqi committed 1 year ago

Verified 1c65ad74

Updates Contributing.md (#8343)

mikegre-google committed 1 year ago

Verified 19c4d632

[Documentation] Added a section pointing readers to the AI-Hypercomputer/tpu-recipies repo for reference model implementations. (#8412)

mikegre-google committed 1 year ago

Verified 8e10036b

Align rrelu_with_noise schema to reflect noise mutation [branch] (#8363)

IvanKobzarev committed 1 year ago

Verified 73f54ba5

[scan] Add a test under SPMD (#8419)

tengyifei committed 1 year ago

Verified 2ba14fcd

Switch to libtpu package from the libtpu-wheels registry (#8409)

tengyifei committed 1 year ago

Verified ae172982

Update `train_resnet_benchmark.py` reference (#8428)

emmanuel-ferdman committed 1 year ago

Verified b66c2494

Reenable the distributed checkpointing test (#8424)

JackCaoG committed 1 year ago

Verified 591c3979

Support SegmentID when doing data prallel SPMD (#8425)

JackCaoG committed 1 year ago

Verified 1c91219f

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

qihqi committed 1 year ago

Verified 20f51661

Update XLA pin to 11/22 (#8411)

JackCaoG committed 1 year ago

Verified 39e67b5d

scan and scan_layers (#7901)

tengyifei committed 1 year ago

Verified 51575db4

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

qihqi committed 1 year ago

Verified 31d348e5

Force specialize_float = True for torch_xla (#8404)

ezyang committed 1 year ago

Verified d503ca54

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

vanbasten23 committed 1 year ago

Verified d572aeb5

Unify how cpp test is being run (#8391)

JackCaoG committed 1 year ago

Verified 2ec22641

Update ARC runner version to 2.321.0 (#8397)

mbzomowski committed 1 year ago

Verified fbed916a

Update XLA pin to 11/17 (#8392)

JackCaoG committed 1 year ago

Verified 0c3d54bf

Fix the fallback op in SPMD (#8386)

JackCaoG committed 1 year ago

Verified defd0a9d

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

ManfeiBai committed 1 year ago

Verified 91f5c8ad

Fix incorrect xla type for view_as_complex (#8382)

JackCaoG committed 1 year ago

Verified bd4006ef

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

JackCaoG committed 1 year ago

Verified 5b3c3315

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

vanbasten23 committed 1 year ago

Verified 102cd489

Expose lazy all number special scalars (#8361)

rpsilva-aws committed 1 year ago

Verified f610f704

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

ManfeiBai committed 1 year ago

Verified d4c90c91

Newer Older