[pull] master from tensorflow:master#1982
Merged
Merged
Conversation
PiperOrigin-RevId: 939972806
…lete * Add AlignedNew and AlignedDelete to tsl::port and implement them across Linux, default, and Windows ports using C++17 aligned operator new/delete. * Add AllocateRawAlignedNew and DeallocateRawAlignedNew virtual overloads to Allocator and override them in CPUAllocator. * Update TypedAllocator to call AllocateRawAlignedNew and DeallocateRawAlignedNew. PiperOrigin-RevId: 939990435
- Replace custom implementations of "ASSIGN_OR_RETURN" and "RETURN_IF_ERROR" with absl's standard implementations - Define with flag = 1 to define shortened aliases to absl's implementations, prefixed with ABSL PiperOrigin-RevId: 940003092
Free the default 0-size `node->temporaries` array before allocating and assigning a new one in `GetRequiredTemporaries`. Previously, this 0-size array (allocated in `Subgraph::AddNodeWithParameters`) was overwritten without being freed, leaking 16 bytes (on arm64) per delegated partition. This affects both `metal_delegate.mm` and `delegate.cc`. PiperOrigin-RevId: 940012449
Remove obsolete code from xla_tpu_device. PiperOrigin-RevId: 940013090
This header is causing build problems on various platforms. PiperOrigin-RevId: 940058593
…inputs When tf.raw_ops.LRN or LRNGrad is compiled with XLA for float16 or bfloat16 inputs, compilation fails with an internal error due to mixed precision operands in binary operations (e.g. multiply). This happens because attribute scalar constants (bias_, alpha_, beta_) were created using xla::ConstantR0<float>, which hardcodes F32 precision. This change replaces xla::ConstantR0<float> with XlaHelpers::FloatLiteral(builder, input_type(0), ...) in LRNOp::Compile and LRNGradOp::Compile so that scalar attribute constants match the precision of the input tensor. PiperOrigin-RevId: 940067706
PiperOrigin-RevId: 940071702
…w progress. This change introduces two warning mechanisms in HloRematerialization: 1. A warning is logged if the rematerialization pass takes longer than 3 minutes. 2. During the PeakPriority algorithm, if a subpass reduces peak memory by less than 1% while still being over the memory limit, a warning is logged. This helps identify cases where rematerialization is making very slow progress. PiperOrigin-RevId: 940075135
…when exporting while op StableHLO allows implicit references to values defined outside of a while loop, which gets appended as explicit operands while being converted to HLO. This updates the original value of the result of such a while loop accordingly by appending elments of the original values of the new operands to the existing original value. PiperOrigin-RevId: 940075435
PiperOrigin-RevId: 940076911
…t heap buffer overflow and out-of-bounds read. PiperOrigin-RevId: 940093007
Remove unused TPU compile proto dependencies from tf2xla APIs. The `tpu_compile.proto.h` and related build targets are no longer required by the tf2xla v1 and v2 APIs, so they are being removed to clean up dependencies. PiperOrigin-RevId: 940095382
This new workflow uses `addlicense` to ensure that all relevant source files in the TensorFlow repository have the correct Apache 2.0 license header. It runs on pull requests and pushes to the main branch, skipping files in `third_party` and various non-source file types. Additionally, Apache 2.0 license headers were added to the following files: - `tensorflow/python/framework/fast_tensor_util.pyx` - `tensorflow/tools/pip_package/xla_build/CMakeLists.txt.oss` - `tensorflow/tools/pip_package/xla_build/pip_test/CMakeLists.txt` PiperOrigin-RevId: 940101317
PiperOrigin-RevId: 940102495
Extract the logic for updating the value set of a single operand of an async-start instruction into a new helper method, UpdateAsyncChainOperandValueSet. This improves modularity and allows for potential reuse of this logic. PiperOrigin-RevId: 940104470
The chlo.topk operation now includes an optional is_stable attribute, defaulting to true. This attribute controls whether the sort performed by topk is stable. The legalization to stablehlo.sort now propagates this is_stable value. PiperOrigin-RevId: 940120099
Updates LLVM usage to match [72c5132a2e52](llvm/llvm-project@72c5132a2e52) PiperOrigin-RevId: 940131349
…k and downcasts. PiperOrigin-RevId: 940132064
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
10 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot] (v2.0.0-alpha.4)
Can you help keep this open source service alive? 💖 Please sponsor : )