Repository navigation
[pull] master from tensorflow:master - #8854
Merged
Merged
Conversation
…conv-rank PiperOrigin-RevId: 988536632
PiperOrigin-RevId: 988569222
The TFLite and LiteRT CMake builds cloned the whole TensorFlow repository at configure time, but only needed a few files from it. - Built converter sources from tflite/converter instead of the TF checkout. - Added Findtensorflow_headers.cmake to download the 6 header-only files (xla/tsl, third_party/fft2d) that the TFLite runtime includes. - Fetched TensorFlow only for benchmark_model and the kernel tests. - Removed TENSORFLOW_SOURCE_DIR usages from LiteRT CMake files. - Removed unused TensorFlow logging include from nnapi_handler, which broke the Android CMake build without the TF checkout. PiperOrigin-RevId: 988570167
…inor for all matmuls so user doesn't have to. PiperOrigin-RevId: 988574639
Breaking change: these messages move to the `xla.megascale.runtime.external` package and drop their `Overlay` suffix, e.g. `xla::megascale::runtime::MegaScaleRuntimeErrorOverlay` is now `xla::megascale::runtime::external::MegaScaleRuntimeError`. The wire format is unchanged. PiperOrigin-RevId: 988595845
PiperOrigin-RevId: 988599619
The CMake benchmark_model build fetched the TensorFlow repository to compile
tf.Example and TFLite profiling protos. Compile the proto-based outputs out
instead, so benchmark_model builds without TensorFlow or protobuf.
- Added the TFLITE_WITHOUT_PROTO define to benchmark_tflite_model.cc and
profile_summary_formatter.{h,cc}, guarding the proto-based outputs
(--output_proto_filepath, --result_file_path, --export_model_runtime_info
and the 'proto' op profiling output mode).
- Set TFLITE_WITHOUT_PROTO for benchmark_model and label_image in CMake and
dropped their proto link dependencies.
- Removed the CMake files for the benchmark result and profiling protos.
- Downloaded stats_calculator and its forwarding headers in
Findtensorflow_headers.cmake instead of taking them from the TensorFlow
checkout.
- Validated the field count of each Findtensorflow_headers.cmake entry.
- Fetched TensorFlow only for the kernel tests.
- Removed the unused TFLITE_GENERATED_HEADERS_DIR variable.
PiperOrigin-RevId: 988641036
PiperOrigin-RevId: 988666007
…ts against Tsplits limits. PiperOrigin-RevId: 988777009
Replace long(2)**32 + 3 with 2**32 + 3 and add the missing test.main() entrypoint. PiperOrigin-RevId: 988876726
PiperOrigin-RevId: 988877018
The CMake kernel tests were the last user of the TensorFlow checkout. They now build from LiteRT sources only. - Took the schema and converter sources from tflite/converter and generated the mutable schema header in the build directory. - Dropped the unused compiler/mlir/tools/optimize/quantization_utils.cc and the tsl logging sources. - Replaced the tsl logging include in kernels/test_util with absl/log. - Added TFLITE_WITHOUT_MODEL_DUMP to compile out test model dumping, which needs tools/serialization, and set it in the CMake build. - Fixed stale sources and link deps (op_version, op_signature, absl check/log/random). - Removed the TensorFlow FetchContent block and TENSORFLOW_SOURCE_DIR. PiperOrigin-RevId: 988908594
…to reject invalid operations with mismatched inner dimensions. Fixes #117771 PiperOrigin-RevId: 988931357
…d HloCloneContext. Replace read-only `const std::vector<T>&` and `const std::string&` parameters with `absl::Span<const T>` and `absl::string_view` in `HloModuleConfig::set_spmd_parameters_shardings`, `HloModule::set_profile_info`, `HloInstruction` cycle detection, and `HloCloneContext` to avoid unnecessary container copies and allow callers to pass slices or views directly. PiperOrigin-RevId: 988979844
…eters. Replace read-only `const std::string&` parameters with `absl::string_view` across `HloModuleMetadata`, `HloPassInterface`, and `HloPassPipeline` to avoid unnecessary `std::string` heap allocations when string literals or views are passed. PiperOrigin-RevId: 989002614
Replace `const std::string&` parameters with `absl::string_view` in `HloInstruction` methods (`CreateInfeed`, `Clone`, `CloneWithNewShape`, `CloneWithNewOperands`, `add_frontend_attribute`, `set_infeed_config`, `set_outfeed_config`) and string-to-enum helpers (`StringToRandomAlgorithm`, `StringToRandomDistribution`, `StringToPrecision`, `StringToAlgorithm`) to avoid unnecessary `std::string` heap allocations. PiperOrigin-RevId: 989037408
Replace iterator-based `std::find`, `std::replace`, `std::all_of`, and manual search loops with `absl::c_linear_search`, `absl::c_find`, `absl::c_any_of`, `absl::c_replace`, and `absl::c_all_of` across `HloInstruction`, `HloInstructionSequence`, and `OriginalValue`. PiperOrigin-RevId: 989057167
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.
6 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 : )