GitHub / pytorch/xla issues and pull requests
#9483 - Update TPU CI with latest docker container
Pull Request -
State: closed - Opened by bhavya01 16 days ago
#9480 - Bump version
Pull Request -
State: open - Opened by qihqi 17 days ago
#9476 - [Kernel] Update ragged attention block table
Pull Request -
State: closed - Opened by yaochengji 20 days ago
#9475 - Fix spmd sharding visualization when device index is >= 10
Pull Request -
State: open - Opened by jeffhataws 20 days ago
#9474 - Add dtensor mesh conversion test
Pull Request -
State: open - Opened by aws-cph 21 days ago
#9474 - Add dtensor mesh conversion test
Pull Request -
State: closed - Opened by aws-cph 21 days ago
- 2 comments
#9473 - Optimize w8a8 pallas kernel
Pull Request -
State: open - Opened by kyuyeunk 21 days ago
#9472 - "RuntimeError: !at::functionalization::impl::isFunctionalTensor(t)" when running a DTensor test with functionalization on
Issue -
State: open - Opened by jeffhataws 21 days ago
- 1 comment
#9471 - Remove duplicate artifact creation for 2.8.0-rc1
Pull Request -
State: closed - Opened by pgmoka 21 days ago
#9470 - Calculate vmem limit dynamically in the quantized matmul kernel.
Pull Request -
State: closed - Opened by vanbasten23 21 days ago
- 1 comment
#9469 - Update cuda version check
Pull Request -
State: closed - Opened by pgmoka 22 days ago
#9468 - By default, to("jax") should go to TPU
Pull Request -
State: closed - Opened by zzzwen 22 days ago
#9467 - feat: abstraction of xla::OpSharding proto using wrapper class
Pull Request -
State: open - Opened by kvshbg-aws 22 days ago
#9466 - Lack of 2.9 wheel causing torchprime test error
Issue -
State: open - Opened by pgmoka 22 days ago
Labels: bug, install
#9465 - Remove the clamp op when we do symmetric quantization on a tensor
Pull Request -
State: closed - Opened by vanbasten23 22 days ago
- 3 comments
#9464 - Error Handling: replace `ConsumeValue` with `GetValueOrThrow`.
Pull Request -
State: open - Opened by ysiraichi 22 days ago
#9463 - Partially disable tpu-info CLI tests
Pull Request -
State: closed - Opened by bhavya01 23 days ago
#9462 - Re-enable tpu-info cli tests
Issue -
State: open - Opened by bhavya01 23 days ago
Labels: bug, libtpu, CI, xla:tpu
#9461 - Change nightly_package_version to 2.9
Pull Request -
State: closed - Opened by pgmoka 23 days ago
- 1 comment
#9460 - Make assume_pure able to work with functions that depends on random
Pull Request -
State: closed - Opened by qihqi 23 days ago
- 1 comment
#9459 - python test_torch.py -v TestTensorDeviceOpsXLA doesn't run any tests
Issue -
State: open - Opened by bhavya01 23 days ago
Labels: bug, testing, CI
#9458 - Add dtensor placement test
Pull Request -
State: closed - Opened by jeffhataws 23 days ago
#9457 - Error Handling: replace `XLA_CHECK_OK()` with status functions.
Pull Request -
State: closed - Opened by ysiraichi 23 days ago
- 1 comment
#9456 - [DO NOT REVIEW] Verify CI.
Pull Request -
State: closed - Opened by vanbasten23 23 days ago
#9455 - Update torch compat version to 2.7.1
Pull Request -
State: closed - Opened by qihqi 23 days ago
#9454 - Unable to build torch/xla
Issue -
State: closed - Opened by mikegre-google 23 days ago
- 3 comments
Labels: bug, build
#9453 - commit
Pull Request -
State: open - Opened by qihqi 23 days ago
#9452 - Unify the return type of w8a8 matmul between fallback and the actual impl.
Pull Request -
State: closed - Opened by vanbasten23 23 days ago
- 1 comment
#9451 - Add support for callable in torchax.interop.JittableModule.functional_call in the first parameter
Pull Request -
State: open - Opened by zmelumian972 24 days ago
- 2 comments
#9450 - Introduce multi-operand collective permute
Pull Request -
State: open - Opened by rpsilva-aws 24 days ago
#9449 - torch_xla.tpu.version() gets stuck occasionally
Issue -
State: open - Opened by yaochengji 25 days ago
- 3 comments
Labels: bug, xla:tpu
#9448 - Suppress C++ stacktrace on `XLA_CHECK*()` calls.
Pull Request -
State: closed - Opened by ysiraichi 25 days ago
#9447 - [RFC] Controller for SPMD+MPMD
Issue -
State: open - Opened by pgmoka 25 days ago
- 2 comments
Labels: distributed, RFC
#9446 - Fix duplicate labels and other docs build warnings
Pull Request -
State: open - Opened by melissawm 28 days ago
- 3 comments
#9445 - Error Handling: refactor `ExecuteComputation` and `ExecuteReplicated` to propagate status.
Pull Request -
State: open - Opened by ysiraichi 28 days ago
- 1 comment
#9444 - Missing python 3.10 wheel for 2.8.0rc1
Issue -
State: closed - Opened by jeffhataws 29 days ago
- 2 comments
Labels: bug, install
#9443 - Fix CPU tests for python 3.12
Pull Request -
State: closed - Opened by bhavya01 30 days ago
#9442 - implement collective all_to_all op
Pull Request -
State: open - Opened by bfolie 30 days ago
- 1 comment
#9441 - Add `WITH_LOCATION` macros for propagating external library errors.
Pull Request -
State: open - Opened by ysiraichi 30 days ago
- 1 comment
#9440 - Fix status source code location logic.
Pull Request -
State: closed - Opened by ysiraichi 30 days ago
#9439 - Convert some XLA_CHECKs to fatal errors.
Pull Request -
State: open - Opened by zhanyong-wan 30 days ago
#9438 - Add JAX dependency for Python 3.12
Pull Request -
State: closed - Opened by tengyifei about 1 month ago
- 2 comments
#9437 - implement collective reduce op
Pull Request -
State: open - Opened by bfolie about 1 month ago
- 1 comment
#9436 - feat: add normalize_tile_assignment function needed for local SPMD
Pull Request -
State: open - Opened by kvshbg-aws about 1 month ago
#9435 - Implement collective gather op
Pull Request -
State: open - Opened by bfolie about 1 month ago
- 1 comment
#9434 - Update TPU CI container image
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
- 1 comment
#9433 - 2.8 backport PR request list
Issue -
State: open - Opened by pgmoka about 1 month ago
Labels: backport2.8
#9432 - Update r2.8 with the changes being done in 2.8
Pull Request -
State: closed - Opened by pgmoka about 1 month ago
- 1 comment
#9431 - Error Handling: propagate status for `ReleaseGilAndTransferData` and `XlaDataToTensors`.
Pull Request -
State: open - Opened by ysiraichi about 1 month ago
- 1 comment
#9430 - [do not review] check ci
Pull Request -
State: closed - Opened by vanbasten23 about 1 month ago
#9429 - Error Handling: refactor `ComputationClient::TransferFromDevice` to propagate status.
Pull Request -
State: open - Opened by ysiraichi about 1 month ago
- 2 comments
#9428 - Support editable install with setuptools>=80.0.0
Pull Request -
State: closed - Opened by tengyifei about 1 month ago
#9427 - Update CI docker images to use 3.12
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
#9426 - Update tests CI for r2.8
Pull Request -
State: closed - Opened by pgmoka about 1 month ago
- 2 comments
#9425 - Add Build Trigger for 2.8-rc1 release
Pull Request -
State: closed - Opened by pgmoka about 1 month ago
- 1 comment
#9424 - Change update_deps script so that latest stable version can be pulled instead of latest nightly
Pull Request -
State: closed - Opened by bfolie about 1 month ago
#9423 - Install libtpu directly instead of torch_xla[tpu]
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
#9422 - Need to branch for v2.8, and update version to v2.9
Issue -
State: closed - Opened by jeffhataws about 1 month ago
- 3 comments
Labels: bug, install
#9421 - Pin update 2025-06-30
Pull Request -
State: closed - Opened by bfolie about 1 month ago
- 3 comments
#9420 - Error Handling: refactor the existing `ComputationClient` implementations to use status QOL functions.
Pull Request -
State: open - Opened by ysiraichi about 1 month ago
- 1 comment
#9419 - Error Handling: refactor the PjRt registry to use status QOL functions.
Pull Request -
State: closed - Opened by ysiraichi about 1 month ago
- 5 comments
#9418 - [RFC] Enable DTensor SPMD APIs with XLA SPMD Backend
Issue -
State: open - Opened by fhaolinaws about 1 month ago
- 10 comments
Labels: enhancement, distributed, RFC
#9417 - Add support for 3.13 builds
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
#9416 - implement diagonal_copy
Pull Request -
State: closed - Opened by Chenyaaang about 1 month ago
#9415 - Pin update to 20250617. After that there is a pallas regression
Pull Request -
State: closed - Opened by qihqi about 1 month ago
- 1 comment
#9414 - Add a test on OOM error
Pull Request -
State: open - Opened by zhanyong-wan about 1 month ago
#9413 - [TORCHAX] Sharded jax tensors support for FlaxNNModule
Issue -
State: open - Opened by vlad-karp about 1 month ago
- 1 comment
Labels: enhancement, torchxla2
#9412 - Optimize w8a8 quantized matmul kernel
Pull Request -
State: closed - Opened by vanbasten23 about 1 month ago
- 2 comments
#9411 - [TORCHAX] Can't allocate random tensors with device='jax'
Issue -
State: open - Opened by vlad-karp about 1 month ago
- 1 comment
Labels: bug, torchxla2
#9410 - Style improvements.
Pull Request -
State: closed - Opened by zhanyong-wan about 1 month ago
#9409 - Update CODEOWNERS
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
#9408 - Add nightly and dev images for python 3.12
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
- 1 comment
#9407 - Initial support for 3.12
Pull Request -
State: closed - Opened by bhavya01 about 1 month ago
- 1 comment
#9406 - `EmbeddingDenseBackward`: Remove `padding_idx` cast to `double`
Pull Request -
State: closed - Opened by unterumarmung about 1 month ago
- 5 comments
#9405 - Cannot mark sharding or print values of a SPMD tensor in a scanned function
Issue -
State: closed - Opened by Topologized about 1 month ago
- 3 comments
Labels: bug
#9404 - adding tol for numeric test of checkpointing
Pull Request -
State: closed - Opened by yaoshiang about 1 month ago
#9403 - Allgather coalescee: Check tuple shape only if return shape is tuple.
Pull Request -
State: closed - Opened by jeffhataws about 1 month ago
- 3 comments
#9402 - Prepare for pytorch tensor impl change in is_contiguous_custom
Pull Request -
State: closed - Opened by laithsakka about 1 month ago
- 4 comments
#9401 - Revert "[torchax]: JittableModule statedict handling"
Pull Request -
State: closed - Opened by qihqi about 1 month ago
- 1 comment
Labels: ci-no-td
#9400 - chore: trigger CI
Pull Request -
State: closed - Opened by yaoshiang about 1 month ago
#9399 - Pins update for week of 6/16/2025 (attempt #2)
Pull Request -
State: closed - Opened by yaoshiang about 1 month ago
- 3 comments
#9398 - Misc changes: default sharding + allow scalar tensor math
Pull Request -
State: open - Opened by qihqi about 1 month ago
#9397 - Update README.md
Pull Request -
State: closed - Opened by shauheen about 1 month ago
#9396 - [wip]ruff inter on torchax
Pull Request -
State: open - Opened by zzzwen about 1 month ago
- 1 comment
#9395 - [build_developer] Fix vision installation command
Pull Request -
State: closed - Opened by tengyifei about 1 month ago
#9394 - Update documentations for scan cache
Pull Request -
State: closed - Opened by iwknow about 1 month ago
#9393 - Update gru.py to use is_fn_pure
Pull Request -
State: closed - Opened by tengyifei about 1 month ago
- 1 comment
#9392 - Unnecessary FP64 cast for `padding_idx` in `EmbeddingDenseBackward`
Issue -
State: closed - Opened by unterumarmung about 1 month ago
- 3 comments
Labels: enhancement, lowering
#9391 - Changes needed for `device_assignment` in `PjRtComputationClient::Compile` to support submeshing/localized spmd
Issue -
State: open - Opened by kvshbg-aws about 1 month ago
#9390 - Create and Expose the `torch_xla::OpSharding` wrapper class instead of `xla::OpSharding` class
Issue -
State: open - Opened by kvshbg-aws about 1 month ago
Labels: enhancement, distributed
#9389 - Normalize `tile_assignment` after constructing the `xla::OpSharding` object
Issue -
State: open - Opened by kvshbg-aws about 1 month ago
Labels: enhancement, distributed
#9388 - Pins update for week of 6/16/2025
Pull Request -
State: open - Opened by yaoshiang about 1 month ago
#9387 - Update pins for week of June 16, 2025
Issue -
State: open - Opened by yaoshiang about 1 month ago
#9386 - Error Handling: refactor `XlaCoordinator` to use status types.
Pull Request -
State: closed - Opened by ysiraichi about 1 month ago
- 5 comments
#9385 - Fix nested stableHLO composite regions
Pull Request -
State: open - Opened by Carlomus about 1 month ago
- 3 comments
#9384 - ErrorHandling: make `GetComputationClient()` return `StatusOr<T>` type.
Pull Request -
State: closed - Opened by ysiraichi about 1 month ago
- 7 comments
#9383 - Include both pytorch and torch_xla revisions in the compilation cache key.
Pull Request -
State: closed - Opened by zhanyong-wan about 1 month ago
#9382 - Add jax_device context manager to control the device target
Pull Request -
State: open - Opened by zzzwen about 1 month ago
#9381 - Deprecate ShapeOfXlaOp in favor of GetShape
Pull Request -
State: closed - Opened by zhanyong-wan about 1 month ago
- 2 comments
#9380 - Wrap def_static to enable warning reporting in static python functions.
Pull Request -
State: closed - Opened by zhanyong-wan about 1 month ago