pytorch/xla

Pull Requests Commits

[Pallas] Improve the user guide (#7625)

alanwaketan committed 1 year ago

Verified 661f1914

SPMD parameter wrapping fix review comment (#7619)

JackCaoG committed 1 year ago

Verified 3ce6ffb2

add tpu example for torchrun on GKE (#7620)

chengcongdu committed 1 year ago

Verified 51f89c85

Update nightly package version to 2.5 (#7623)

bhavya01 committed 1 year ago

Verified bbf0db79

added how to tell when ops need to be lowered (#7624)

nhira committed 1 year ago

Verified 2c011379

Update ansible config to separately update submodules when torch_pin is specified (#7617)

bhavya01 committed 1 year ago

Verified 59c93619

Handle the parameter wrapping for SPMD (#7604)

JackCaoG committed 1 year ago

Verified 4b7e518d

Lower empty_like, empty_permute, empty_strided (#7610)

lsy323 committed 1 year ago

Verified 1ec9f2bb

Fix GPU yaml in kubernetes.md (#7612)

will-cromar committed 1 year ago

Verified 7ec57784

[torch_xla2] Add lowerings for `rand_like` and `randn_like`, and enable some opinfo tests (#7170)

will-cromar committed 1 year ago

Verified 8f5de773

Run paged_attention test with megacore mode only on TPU v4 (#7609)

bhavya01 committed 1 year ago

Verified e8cd9855

Reland: `mul`: convert inputs to result type. (#7602)

ysiraichi committed 1 year ago

Verified 77f66b51

[benchmarks] Fix `moco` initialization device. (#7598)

ysiraichi committed 1 year ago

Verified 86169d27

Add documentation for Triton kernels in PTXLA (#7595)

bhavya01 committed 1 year ago

Verified e7a8700e

Update JAX version to fix TPU CI (#7601)

will-cromar committed 1 year ago

Verified 593dce8b

add spmd gpu doc back (#7592)

JackCaoG committed 1 year ago

Verified 755ae846

Update OpenXLA-pin to Jun27 (#7334)

ManfeiBai committed 1 year ago

Verified 9c32c66f

get_memory_info can handle default arg (#7590)

JackCaoG committed 1 year ago

Verified 020f3f5d

Let TPU CI read from project GCS buckets (#7600)

will-cromar committed 1 year ago

Verified 6873d332

Update pallas doc with paged_attention (#7591)

wonjoo-wj committed 1 year ago

Verified 3bdca4d1

should_donate should not propagate to the output data (#7585)

JackCaoG committed 1 year ago

Verified 6410989c

Add documentation on fixing a torchbench model on torchxla2 (#7583)

qihqi committed 1 year ago

Verified 6d375f2c

Use upstream trilinear decomp by default (#7586)

JackCaoG committed 1 year ago

Verified 5e51ca23

Add an example of using GKE with torchrun (#7589)

will-cromar committed 1 year ago

Verified 77f69434

Remove references to old release project (#7587)

will-cromar committed 1 year ago

Verified d66a67b5

remove XLA_USE_BF16 and other variants (#7582)

JackCaoG committed 1 year ago

Verified 7d41035b

Clean up experimental (#7579)

JackCaoG committed 1 year ago

Verified ed15ea1e

Add `jlibrary` module to torch_xla2 for preserving library abstractions during export (#7128)

GleasonK committed 1 year ago

Verified 91d462cc

[torch_xla2] Prototype of `torch.distributed` backend (#7311)

will-cromar committed 1 year ago

Verified 447031e6

Enable eager spmd (#7341)

JackCaoG committed 1 year ago

Verified d5e57135

Newer Older