Search results

Triton test failed in CI: ImportError: cannot import name 'AttrsDescriptor'

Status: Open.

#8820 In pytorch/xla;

· tengyifei opened on Mar 12, 2025

CollectivePermute support

Status: Open.

#8815 In pytorch/xla;

· rpsilva-aws opened on Mar 11, 2025

Training on multi-slice v6e discovers MegaScale twice again

bugSomething isn't workingSomething isn't working

Status: Open.

#8812 In pytorch/xla;

· tengyifei opened on Mar 10, 2025

xla device precision issue

pytorch divergenceXLA behavior doesn't match Pytorch eager frontendXLA behavior doesn't match Pytorch eager frontend

Status: Open.

#8811 In pytorch/xla;

· mars1248 opened on Mar 10, 2025

MarkShardingFunction causes OOM when applied to model parameters

performance

Status: Open.

#8809 In pytorch/xla;

· tengyifei opened on Mar 8, 2025

Use jax autograd from PyTorch

enhancementNew feature or requestNew feature or request

torchxla2

Status: Open.

#8805 In pytorch/xla;

· tengyifei opened on Mar 7, 2025

Re-enable CPU test test/test_python_ops.py -k TestPythonOps for uint8 dtype

bugSomething isn't workingSomething isn't working

libtpu

Status: Open.

#8799 In pytorch/xla;

· tengyifei opened on Mar 5, 2025

xb.Op.scalar maps unsigned int to torch signed int

bugSomething isn't workingSomething isn't working

good first issueGood for newcomersGood for newcomers

needs reproduction

Status: Open.

#8798 In pytorch/xla;

· tengyifei opened on Mar 5, 2025

Test flake on GPU: CUDA_ERROR_ILLEGAL_ADDRESS

CICI related changeCI related change

testingTesting and coverage related issues.Testing and coverage related issues.

xla:gpu

Status: Open.

#8796 In pytorch/xla;

· tengyifei opened on Mar 5, 2025

xb.call_jax only works with torch.Tensor arguments

enhancementNew feature or requestNew feature or request

Status: Open.

#8795 In pytorch/xla;

· tengyifei opened on Mar 5, 2025

xb.call_jax always assumes the traced function makes use of all arguments

enhancementNew feature or requestNew feature or request

Status: Open.

#8794 In pytorch/xla;

· tengyifei opened on Mar 5, 2025

After importing jax first, any PyTorch/XLA computation will hang

bugSomething isn't workingSomething isn't working

needs reproduction

Status: Open.

#8793 In pytorch/xla;

· tengyifei opened on Mar 5, 2025