pytorch/xla

Pull Requests Commits

Allow MpDeviceLoader to shard dictionaries of tensor for 2.5 release

bhavya01 committed 1 year ago

58cdbcce

change r2.5 to use jax stable version 0.4.33 (#8062)

ManfeiBai committed 1 year ago

Verified c074e313

[Cherry-pick] Enable cross entropy loss for xla autocast with FP32 precision (#8094) (#8201)

ManfeiBai committed 1 year ago

Verified 31a0a927

[cherry-pick] Update Openxla-pin to Sep13, libtpu-pin to Sep13, jax to 0.4.33 stable version (#8052)

ManfeiBai committed 1 year ago

Verified 3bb8736f

Force github workflow catch PyTorch commit from release/2.5 branch for PyTorch/XLA `r2.5` branch (#8078)

ManfeiBai committed 1 year ago

Verified 5cc01cf1

cherry pick collective op all_to_all_single (#8077)

zpcore committed 1 year ago

Verified 8f43eef8

Use Tuple instead of tuple in scan.py (#8067)

tengyifei committed 1 year ago

Verified 1717e258

Set xla_tpu_enable_flash_attention=false to enable libtpu pin update (#8008)

bhavya01 committed 1 year ago

3c7daa24

Follow 2 releases grace period for API deprecation (#8003)

zpcore committed 1 year ago

ba13b4ea

Add torch_pin to 2.5 release branch

ManfeiBai committed 1 year ago

Verified 87304aa0

[torch_ xla2] LLama3 example - fixed sharding configuration on LLama3 example (#7973)

zmelumian committed 1 year ago

Verified ef6b3703

[torch_xla2] Fix the test of `cumsum` (#7965)

guyao committed 1 year ago

Verified 12e59584

Update Neuron initializations (#7952)

jeffhataws committed 1 year ago

Verified 901c3a32

grid_sampler_3d, gaussian_nll_loss, dropout (#7969)

qihqi committed 1 year ago

Verified 900296aa

Remove `TPU_LIBRARY_PATH` from legacy TPU CI (#7968)

will-cromar committed 1 year ago

Verified 0528c9d7

Increment version in upstream CI `docker push` command (#7967)

will-cromar committed 1 year ago

Verified c201ec11

Revert "Allow MpDeviceLoader to shard dictionaries of tensors" (#7964)

JackCaoG committed 1 year ago

Verified 59570c77

Unpin `scipy` in upstream CI image (#7963)

will-cromar committed 1 year ago

Verified 282bf932

Run ansible release stage after local builds are installed (#7955)

will-cromar committed 1 year ago

Verified 8bebeddd

Remove numpy pin for upstream image (#7954)

will-cromar committed 1 year ago

Verified 92f8cfdf

Op info test for logaddexp, logaddexp2 and logcumsumexp (#7506) (#7957)

yenkwang committed 1 year ago

Verified 1ae32bf9

Update upstream docker image to Python 3.10 (#7953)

will-cromar committed 1 year ago

Verified 1a417b4b

Op info test for special.modified_bessel_i1 .. special.polygamma (#7928)

ayshiff committed 1 year ago

Verified e7653b85

Allow MpDeviceLoader to shard dictionaries of tensors (#7912)

bhavya01 committed 1 year ago

Verified 5acecc33

`DynamicShapeDetector` with trie implementation. (#7918)

ysiraichi committed 1 year ago

Verified 400bd91b

support dist.reduce_scatter_tensor (#7950)

zpcore committed 1 year ago

Verified b1fbb749

Skip dynamo collective tests on TPU v2 CI (#7949)

will-cromar committed 1 year ago

Verified f45775a5

bring XLA_USE_BF16 and XLA_DOWNCAST_BF16 back for 2.5 release (#7945)

JackCaoG committed 1 year ago

Verified d224c7a3

Include Terraform lock for release triggers (#7948)

will-cromar committed 1 year ago

Verified 2540ae78

Ignore torch_xla2 on legacy TPU CI (#7947)

will-cromar committed 1 year ago

Verified 3a64bbc8

Older