Repository navigation

[pull] master from tensorflow:master - #8857

Merged

pull[bot] merged 21 commits into

Cache-Cloud:masterfrom

tensorflow:master

Sep 29, 2026

Merged

[pull] master from tensorflow:master#8857
pull[bot] merged 21 commits into
Cache-Cloud:masterfrom
tensorflow:master

Commits on Sep 29, 2026
- [XLA:GPU] Add asymmetric mode to BlockBarrierOp
  
  sohaibiftikhar
  authored and
  tensorflower-gardener
  committed
  25453d5
- Deallocate GPU buffers in TopKKernelTest to prevent OOM across parameterized runs
  
  IllogicalMoose
  authored and
  tensorflower-gardener
  committed
  e2957f2
- Build the TFLite C library from LiteRT sources in CMake
  
  terryheo
  authored and
  tensorflower-gardener
  committed
  7b8368c
- [XLA:GPU] Add a test for a tiled reduce duplicated across concatenate regions.
  
  Moerafaat
  authored and
  tensorflower-gardener
  committed
  46a6166
- Configure DTensor GPU test defaults for Ampere+ GPUs and smaller GPU memory partitions
  
  IllogicalMoose
  authored and
  tensorflower-gardener
  committed
  dde838b
- Update deprecated MLIR API call sites
  
  mrguenther
  authored and
  tensorflower-gardener
  committed
  53de455
- Offload EagerContext destructor to a detached thread when it is the last reference.
  tensorflower-gardener
  committed
  463eb9e
- Replace hardcoded requires-gpu-nvidia tags in tf_cuda_cc_test targets
  
  IllogicalMoose
  authored and
  tensorflower-gardener
  committed
  403dcdb
- `ConcurrentRegionsHloOrdering::CanRunConcurrently` previously rejected all `HloOpcode::kCustomCall` instructions, isolating every typed FFI custom kernel into its own singleton concurrent region wi…
  
  iumitakgun
  authored and
  tensorflower-gardener
  committed
  0b51168
- Account for Hopper reserved shared memory in CudaRuntimeKernelRegistryTest.ExtractsFuncAttributes
  
  IllogicalMoose
  authored and
  tensorflower-gardener
  committed
  a4b4c6f
- Add async support for onednn on aarch64 (the changes are integrated from https://github.com/openxla/xla/pull/48616)
  tensorflower-gardener
  committed
  eebea9b
- Relax float16 tolerance to 1e-2 in StatefulScatterNdTest._VariableRankTest
  
  IllogicalMoose
  authored and
  tensorflower-gardener
  committed
  dabe608
- Reverts changelist 988511947
  tensorflower-gardener
  committed
  4d007fe
- Disable TensorFloat-32 in JacobianTest.test_grad_jacobian_conv
  
  IllogicalMoose
  authored and
  tensorflower-gardener
  committed
  e20101c
- Remove LoadedExecutable::addressable_devices() and migrate callers to devices().
  
  junwhanahn
  authored and
  tensorflower-gardener
  committed
  4d7aef9
- internal visibility changes
  
  junwhanahn
  authored and
  tensorflower-gardener
  committed
  c1643f8
- [XLA] (2/N User-specified async compute-on) Rewrite compute-on-start/done custom call pairs into async instructions.
  
  seherellis
  authored and
  tensorflower-gardener
  committed
  7b01456
- [XLA:GPU] Record one-shot collective kernels into CUDA graphs with explicit nodes.
  
  PatriosTheGreat
  authored and
  tensorflower-gardener
  committed
  fabfe2a
- Update PjRtRawClient apis to use local_device_id instead of PjRtDevice and
  
  pschuh
  authored and
  tensorflower-gardener
  committed
  0086dd6
- Remove LoadedExecutable::num_devices() and migrate callers to devices().
  
  junwhanahn
  authored and
  tensorflower-gardener
  committed
  3cdbc24
- Add TfLiteBlockwiseQuantizationV2 for block_shape in the TFLite C API.
  
  fengwuyao
  authored and
  tensorflower-gardener
  committed
  96138a6

Commits

Commits on Sep 29, 2026