-
Notifications
You must be signed in to change notification settings - Fork 545
Insights: pytorch/xla
Overview
Could not load contribution data
Please try again later
18 Pull requests merged by 13 people
-
Allgather coalescee: Check tuple shape only if return shape is tuple.
#9403 merged
Jun 25, 2025 -
Prepare for pytorch tensor impl change in is_contiguous_custom
#9402 merged
Jun 25, 2025 -
ErrorHandling: make
GetComputationClient()returnStatusOr<T>type.#9384 merged
Jun 24, 2025 -
Migrate to correct logger interface
#9191 merged
Jun 24, 2025 -
Support torch.distributed.scatter collective
#9365 merged
Jun 24, 2025 -
Revert "[torchax]: JittableModule statedict handling"
#9401 merged
Jun 23, 2025 -
[torchax]: JittableModule statedict handling
#9195 merged
Jun 22, 2025 -
Update README.md
#9397 merged
Jun 20, 2025 -
Add jax_device context manager to control the device target
#9382 merged
Jun 20, 2025 -
Refactor call_jax to allow implementing ops in python
#9354 merged
Jun 20, 2025 -
[build_developer] Fix vision installation command
#9395 merged
Jun 19, 2025 -
Update gru.py to use is_fn_pure
#9393 merged
Jun 19, 2025 -
Update documentations for scan cache
#9394 merged
Jun 19, 2025 -
Make the shape property adhere to torch.Tensor interface.
#9342 merged
Jun 19, 2025 -
Torchax: JittableModule isinstance to work with encapsulated model
#9375 merged
Jun 19, 2025 -
Add cache support for scan_layers
#9297 merged
Jun 18, 2025 -
Wrap def_static to enable warning reporting in static python functions.
#9380 merged
Jun 18, 2025 -
Revert "Avoid unnecessary copy in TensorSource (#8849)"
#9379 merged
Jun 18, 2025
6 Pull requests opened by 6 people
-
Fix nested stableHLO composite regions
#9385 opened
Jun 18, 2025 -
Error Handling: refactor `XlaCoordinator` to use status types.
#9386 opened
Jun 18, 2025 -
[wip]ruff inter on torchax
#9396 opened
Jun 20, 2025 -
Misc changes: default sharding + allow scalar tensor math
#9398 opened
Jun 22, 2025 -
adding tol for numeric test of checkpointing
#9404 opened
Jun 25, 2025 -
EmbeddingDenseBackward: Remove `padding_idx` cast to `double`
#9406 opened
Jun 25, 2025
2 Issues closed by 2 people
-
Enable torch_xla::OpSharding to establish the relationships for xla::DeviceAssignment
#9357 closed
Jun 24, 2025 -
[scan] Avoid re-tracing the combine function on every call
#8632 closed
Jun 19, 2025
6 Issues opened by 4 people
-
Cannot mark sharding or print values of a SPMD tensor in a scanned function
#9405 opened
Jun 25, 2025 -
Unnecessary FP64 cast for `padding_idx` in `EmbeddingDenseBackward`
#9392 opened
Jun 18, 2025 -
Create and Expose the `torch_xla::OpSharding` wrapper class instead of `xla::OpSharding` class
#9390 opened
Jun 18, 2025 -
Normalize `tile_assignment` after constructing the `xla::OpSharding` object
#9389 opened
Jun 18, 2025 -
Update pins for week of June 16, 2025
#9387 opened
Jun 18, 2025
17 Unresolved conversations
Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.
-
implement send and recv using collective_permute
#9373 commented on
Jun 20, 2025 • 6 new comments -
wip: use torch from a wheel
#9340 commented on
Jun 18, 2025 • 0 new comments -
Check the scan op input for requires_grad
#9083 commented on
Jun 20, 2025 • 0 new comments -
Update test_export_fx_passes.py
#8933 commented on
Jun 18, 2025 • 0 new comments -
[benchmarks] Add docs file.
#8023 commented on
Jun 24, 2025 • 0 new comments -
Shardy support
#9348 commented on
Jun 25, 2025 • 0 new comments -
TPU test flake: SIGSEGV in train_decoder_only_eager_spmd_data_parallel.py
#9046 commented on
Jun 25, 2025 • 0 new comments -
Failing `torch_xla._XLAC._xla_custom_call()` with `RuntimeError: Bad StatusOr access: UNIMPLEMENTED: No registered implementation for custom call to my_lib.my_op.default for platform CUDA`
#9371 commented on
Jun 24, 2025 • 0 new comments -
Leverage torch_xla::OpSharding in mark_sharding
#9358 commented on
Jun 24, 2025 • 0 new comments -
[torch-xla v2.8] Error "Check failed: state Expected an array shape." when running test/test_mp_reduce_scatter.py
#9314 commented on
Jun 24, 2025 • 0 new comments -
Enable lazy tensor loading for sharded tensors
#9341 commented on
Jun 23, 2025 • 0 new comments -
PyTorch/XLA custom Triton kernel export to StableHLO
#9366 commented on
Jun 23, 2025 • 0 new comments -
Support `torch.Generator` on PyTorch/XLA
#9159 commented on
Jun 23, 2025 • 0 new comments -
Torchax should be a separate wheel or a submodule
#9363 commented on
Jun 23, 2025 • 0 new comments -
[v2.8] Neuron inference trace analyzer/bucketing unit tests hanging at GetParameterIdTensorMapping/TransferFromDevice
#9378 commented on
Jun 23, 2025 • 0 new comments -
[RFC] Improved coverage for native distributed collective operations
#9315 commented on
Jun 20, 2025 • 0 new comments -
normalize error handling in PyTorch/XLA
#9339 commented on
Jun 20, 2025 • 0 new comments