pytorch/xla

Pull Requests Commits

[LoweringContext] SPMD propagation (#8471)

rpsilva-aws committed 1 year ago

Verified b068cab9

Modify the deprecation plan for XLA_USE_BF16 (#8474)

rpsilva-aws committed 1 year ago

Verified 18128175

Fix #8475: add libtpu-wheels registry when building docker (#8476)

tengyifei committed 1 year ago

Verified 47f737a7

Add autocast support for einsum (#8420)

aws-nm9 committed 1 year ago

Verified 7b5aca68

Remove Jack and Will from code owner (#8473)

JackCaoG committed 1 year ago

Verified 5c062ea3

Support S32/U32 indices for BWD embedding & Neuron implicit downcast (#8462)

rpsilva-aws committed 1 year ago

Verified 00c0e966

[LoweringContext] Support an optimized parameter mapping for SPMD (#8460)

rpsilva-aws committed 1 year ago

Verified 5d11f660

Improve wheel naming logic in ansible (#8465)

tengyifei committed 1 year ago

Verified cff26e5b

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

Newer Older