Weekly GitHub Report for Tensorflow: May 05, 2025 - May 12, 2025 (12:00:01)
Weekly GitHub Report for Tensorflow
Thank you for subscribing to our weekly newsletter! Each week, we deliver a comprehensive summary of your GitHub project's latest activity right to your inbox, including an overview of your project's issues, pull requests, contributors, and commit activity.
Table of Contents
I. News
1.1 Recent Version Releases:
The current version of this repository is v2.19.0
1.2 Version Information:
The TensorFlow 2.19.0 release, created on March 5, 2025, introduces breaking changes in the LiteRT (tf.lite) C++ and Python APIs, including the transition of tf.lite.Interpreter to a new location with a deprecation warning. Notable improvements include support for bfloat16 in the tfl.Cast operation, and the discontinuation of publishing libtensorflow packages, which can still be accessed via PyPI.
II. Issues
2.1 Top 5 Active Issues:
We consider active issues to be issues that that have been commented on most frequently within the last week. Bot comments are omitted.
-
clang: error: unknown argument: '-fno-canonical-system-headers': This issue involves a build error encountered when attempting to compile TensorFlow using Spack, specifically related to the argument '-fno-canonical-system-headers' which is not recognized by clang. The user is trying to build TensorFlow with CUDA support on a RHEL9 x86_64 platform using various versions of TensorFlow and compilers, but encounters the same error with both clang and gcc.
- The comments discuss a related issue on Spack's GitHub, question the safety of ignoring the '-fno-canonical-system-headers' argument, and note that TensorFlow depends on the 'libclang' Python package from version 2.7. Additionally, a snippet from TensorFlow's configuration files is shared, explaining the rationale behind using the '-fno-canonical-system-headers' flag for gcc.
- Number of comments this week: 4
-
tf.math.igammadoesn't work when dtype is bfloat16 or half: This issue reports a bug in TensorFlow'stf.math.igammafunction, which crashes when the input data type is set tobfloat16orhalf, despite the documentation stating these types are supported. The problem does not occur when usingfloat32orfloat64, and attempts to use XLA JIT compilation do not resolve the issue.- The comments highlight that similar issues exist with
tf.math.igammacandtf.math.lgamma, and mention thattf.math.rintalso fails withbfloat16, suggesting a broader problem with these data types across multiple functions. - Number of comments this week: 3
- The comments highlight that similar issues exist with
-
tf.truncatedivdoesn't work on complex64 and complex128: This issue reports a bug in TensorFlow where thetf.truncatedivfunction does not work with complex number data typescomplex64andcomplex128, despite the documentation indicating that these types should be supported. The problem is demonstrated with a code snippet that attempts to perform a division operation usingtf.truncatedivon complex numbers, resulting in aNotFoundError.- The comment section contains a single comment that does not provide any additional information or context about the issue.
- Number of comments this week: 1
-
[iOS 18] EXC_BAD_ACCESS when initializing TFLite model on device running iOS 18.4.1: This issue reports a bug where an application crashes with an EXC_BAD_ACCESS error when initializing a TensorFlow Lite model on a device running iOS 18.4.1, while the same setup works without issues on iOS 17.x. The problem is specific to the initialization of the TFLite interpreter, and the user is seeking confirmation on whether this is due to an incompatibility with iOS 18 or if a fix is being developed.
- The user is inquiring about potential incompatibilities with iOS 18 and is asking if there is a fix in progress for the issue.
- Number of comments this week: 1
Since there were fewer than 5 open issues, all of the open issues have been listed above.
2.2 Top 5 Stale Issues:
We consider stale issues to be issues that has had no activity within the last 30 days. The team should work together to get these issues resolved and closed as soon as possible.
SystemErrorintf.ensure_shapeandtf.compat.v1.ensure_shapewhendtypeofshapeistf.uint64and its value is too large.: This issue involves a bug in TensorFlow where usingtf.ensure_shapeortf.compat.v1.ensure_shapewith ashapeofdtypetf.uint64and a value close to 2^64 results in aSystemErrorandOverflowError. The problem has been reproduced with TensorFlow Nightly on a Linux Ubuntu 20.04 system using Python 3.10, and it specifically occurs when the shape value is extremely large, such as[18446743219011059112, 1].- Feature Request: Integrate different Digital Signal Processing into tf.signal: This issue is a feature request to integrate advanced Digital Signal Processing (DSP) functionalities into TensorFlow's
tf.signalmodule, similar to those available in the PyTorch ecosystem, particularly the julius library. The request highlights the need for native tools within TensorFlow to perform complex audio data augmentation, which would enhance the capabilities of researchers and developers working on audio-related projects by streamlining workflows and reducing dependency on external libraries. - [DOCS] Missing complex input for Round op: This issue highlights a documentation bug in TensorFlow's
Roundoperation, where the official documentation incorrectly states that a complex tensor can be used as input, but in practice, this does not work as expected. The user reports that they must apply theRoundoperation separately to the real and imaginary parts of the tensor to achieve the desired result, indicating a discrepancy between the documented functionality and the actual behavior in TensorFlow version 2.15.0. - tf.raw_ops.Unbatch aborts with "Check failed: d < dims()": This issue pertains to a bug in TensorFlow version 2.17, where the
tf.raw_ops.Unbatchoperation aborts unexpectedly with an error message "Check failed: d < dims()". The problem has been reproduced using TensorFlow Nightly on a Linux Ubuntu 20.04.3 LTS system with Python 3.11.8, and involves custom code that triggers the error when attempting to unbatch a tensor. Since there were fewer than 5 open issues, all of the open issues have been listed above.
2.3 Open Issues
This section lists, groups, and then summarizes issues that were created within the last week in the repository.
Issues Opened This Week: 9
Summarized Issues:
- TensorFlow GPU Inference Discrepancies: This issue highlights a bug where running inference on a simple Keras model using TensorFlow on macOS M1 yields different results when executed on a GPU compared to a CPU. Consistent results are only achieved when the GPU is not visible to TensorFlow, indicating a potential problem with GPU execution.
- TensorFlow Lite GPU Compatibility: This issue addresses the compatibility challenges faced when using TensorFlow Lite with GPU delegation for transformer models like MobileBERT. Problems with batch size handling during inference are highlighted, where the GPU delegate fails due to unsupported batch sizes and misinterpretation of tensor dimensions.
- TensorFlow Function Data Type Support Bugs: Several issues describe bugs in TensorFlow functions where the
tf.math.igamma,tf.math.reciprocal, andtf.truncatedivfunctions fail to support certain data types as documented. These bugs result in crashes or errors when usingbfloat16,half, integer types, and complex data types, despite documentation indicating support.
- TensorFlow Lite iOS Compatibility Issue: This issue involves a bug where an application crashes with an EXC_BAD_ACCESS error when initializing a TensorFlow Lite model on iOS 18.4.1. The same model and code function correctly on iOS 17.x, indicating a potential compatibility problem with the newer iOS version.
- TensorFlow Build Errors: Multiple issues involve build errors encountered while compiling TensorFlow or TensorFlow Lite. Errors include unsupported AVX512BF instructions and unknown compiler flags, indicating problems with configuration and compatibility of certain compiler settings.
- TensorFlow Documentation Discrepancies: This issue highlights a discrepancy between the TensorFlow documentation and its actual implementation for the
tf.strided_slicefunction. The documentation specifies certain parameter types, but the implementation accepts additional types, indicating a need for documentation updates.
2.4 Closed Issues
This section lists, groups, and then summarizes issues that were closed within the last week in the repository. This section also links the associated pull requests if applicable.
Issues Closed This Week: 12
Summarized Issues:
- DLL Load Errors on Windows: Several issues report DLL load errors when importing TensorFlow on Windows systems, often related to
_pywrap_tensorflow_internal. These errors are attributed to factors like missing MSVC redistributables, unsupported AVX2 instructions, or specific IDE environments like Spyder.
- TensorFlow Installation and Compatibility Issues: Users face challenges installing TensorFlow on various platforms, including macOS with Apple Silicon and Windows, due to compatibility issues. These problems often arise from missing compatible versions or library conflicts with other software like numpy and pandas.
- Model Saving and Attribute Errors: An issue in TensorFlow 2.8 and later versions involves an
AttributeErrorwhen saving models in Google Colab, due to missing attributes in theVariableobject. This problem does not occur in version 2.12, indicating a regression in later versions.
- Compilation and Header File Errors: A compilation error occurs due to a missing header file, "xla/tsl/c/tsl_status.h," which results in a fatal error during the build process of a TensorFlow C program. This indicates a potential issue with the build configuration or missing dependencies.
- Documentation and Deprecation Warnings: A documentation bug highlights a deprecated CMake function,
FetchContent_Populate(abseil-cpp), which should be replaced withFetchContent_MakeAvailable(abseil-cpp). Users are advised to set Policy CMP0169 to OLD temporarily to avoid future issues.
- Function Misunderstandings and Clarifications: A user initially reported a bug in the
tf.math.imagfunction, believing it did not support float and double types. However, it was clarified that the issue was due to an incorrect conversion attempt, not a limitation of the function.
- Community Engagement and Contributions: Steven Lai, a developer from Hong Kong, expressed gratitude for TensorFlow's open-source work and offered his skills for collaboration. He also acknowledged the challenges posed by geographical bias in the tech industry.
- Unspecified or Empty Issues: A closed issue titled "check" lacks context or description, as both the body and comments are empty, providing no additional information about the problem or resolution.
2.5 Issue Discussion Insights
This section will analyze the tone and sentiment of discussions within this project's open and closed issues that occurred within the past week. It aims to identify potentially heated exchanges and to maintain a constructive project environment.
Based on our analysis, there are no instances of toxic discussions in the project's open or closed issues from the past week.
III. Pull Requests
3.1 Open Pull Requests
This section provides a summary of pull requests that were opened in the repository over the past week. The top three pull requests with the highest number of commits are highlighted as 'key' pull requests. Other pull requests are grouped based on similar characteristics for easier analysis. Up to 25 pull requests are displayed in this section, while any remaining pull requests beyond this limit are omitted for brevity.
Pull Requests Opened This Week: 7
Key Open Pull Requests
1. feat: add datatype support for add, ceil, mul, range, sign, sub: This pull request introduces support for additional data types, including bf16, f16, i8, i16, and i32, across various TensorFlow Lite operations such as add, ceil, mul, range, sign, and sub, along with corresponding unit tests and optimizations, while also addressing broadcasting checks and resolving a CONV error with the EIGEN_TFLITE flag.
- URL: pull/92880
- Merged: No
2. Use utility functions to check number of inputs and outputs.: This pull request aims to enhance code efficiency and maintainability in the TensorFlow project by utilizing existing utility functions to verify the number of inputs and outputs, thereby reducing code duplication.
- URL: pull/92832
- Merged: No
3. "Fixed a typo and added a link to official TensorFlow installation guide in README.md"Fix typo and add helpful link to TensorFlow docs: This pull request addresses a minor documentation update by fixing a typo and adding a link to the official TensorFlow installation guide in the README.md file of the TensorFlow project, as detailed in the commit messages and accessible via the provided GitHub URL.
- URL: pull/93101
- Merged: No
Other Open Pull Requests
- Support for f16 and bf16 data types in TensorFlow Lite operations: This topic covers the enhancement of TensorFlow Lite by adding support for f16 and bf16 data types to various elementwise operations. The pull requests ensure that operations like abs, cos, sin, log, rsqrt, sqrt, and square, as well as the exponential function, can now utilize these data types, improving the library's versatility and performance.
- Correction in TOSA component: This pull request addresses a specific issue in the TOSA component of TensorFlow by correcting the use of
obj.isatomlir::isawithin an assertion. This change ensures that the code adheres to the correct usage patterns and improves the reliability of the TOSA component.
- Addition of ASReLU activation function: A new activation function, ASReLU, which combines Swish and ReLU functions with a default alpha value of 1.0, is proposed for inclusion in the TensorFlow project. This addition aims to enhance the project's functionality by providing a new option for activation functions in the
updated.pyfile.
3.2 Closed Pull Requests
This section provides a summary of pull requests that were closed in the repository over the past week. The top three pull requests with the highest number of commits are highlighted as 'key' pull requests. Other pull requests are grouped based on similar characteristics for easier analysis. Up to 25 pull requests are displayed in this section, while any remaining pull requests beyond this limit are omitted for brevity.
Pull Requests Closed This Week: 9
Key Closed Pull Requests
1. fix(WeakTrackableReference): correct weak reference type handling: This pull request addresses the issue of incorrect handling of weak reference types in the WeakTrackableReference class by implementing a more logical approach that checks if the input reference parameter is already a weak reference, converting it if necessary, and using it directly if it is already a weak reference.
- URL: pull/92336
- Merged: 2025-05-07T07:01:00Z
- Associated Commits: 7b8f5
2. Create CsvDataset.py: This pull request aims to enhance TensorFlow's CSV parsing functionality by modifying the 'parse_csv_line' function to support the 'int64' data type alongside the default 'float32', allowing for dynamic handling of columns with different types and enabling the framework to work with datasets containing large integer values.
- URL: pull/92522
- Merged: No
- Associated Commits: cfc93
3. Fixed ppc64le build issues: This pull request addresses and resolves build issues related to oneDNN on the ppc64le platform and fixes a linker error involving an unknown argument '-z defs' with clang-17, as part of the TensorFlow project.
- URL: pull/92769
- Merged: 2025-05-08T05:59:55Z
- Associated Commits: 21be9
Other Closed Pull Requests
- TensorFlow Initialization Testing Enhancements: This topic covers improvements in testing and verification of TensorFlow's initialization logic. The pull request enabled previously disabled tests, added logging for tracking, and improved code readability through comments and formatting.
- MLIR TOSA Component Updates: This topic involves updates to the MLIR TOSA component in TensorFlow to address build errors and naming conventions. The pull requests updated code to accommodate changes in the GreedyRewriteConfig API and aligned naming conventions with the version 1.0 specification.
- Heap-Buffer-Overflow Vulnerability Fix: This topic addresses a potential heap-buffer-overflow vulnerability in the
stringprintf_fuzzfuzzer. The pull request updated the code to include input validation and prevent undefined behavior.
tf.strided_sliceDocumentation and Implementation Alignment: This topic focuses on aligning the implementation and documentation oftf.strided_slicein TensorFlow. The pull request updated thearray_ops.pyfile to ensure consistency as highlighted in issue #92955.
3.3 Pull Request Discussion Insights
This section will analyze the tone and sentiment of discussions within this project's open and closed pull requests that occurred within the past week. It aims to identify potentially heated exchanges and to maintain a constructive project environment.
Based on our analysis, there are no instances of toxic discussions in the project's open or closed pull requests from the past week.
IV. Contributors
4.1 Contributors
Active Contributors:
We consider an active contributor in this project to be any contributor who has made at least 1 commit, opened at least 1 issue, created at least 1 pull request, or made more than 2 comments in the last month.
If there are more than 10 active contributors, the list is truncated to the top 10 based on contribution metrics for better clarity.
| Contributor | Commits | Pull Requests | Issues | Comments |
|---|---|---|---|---|
| mtrofin | 10 | 2 | 0 | 0 |
| RahulSundarMCW | 10 | 2 | 0 | 0 |
| tensorflower-gardener | 11 | 0 | 0 | 0 |
| Venkat6871 | 1 | 0 | 0 | 10 |
| Redempt1onzzZZ | 1 | 1 | 5 | 4 |
| mihaimaruseac | 1 | 0 | 0 | 6 |
| wangw-1991 | 3 | 1 | 1 | 1 |
| 372046933 | 4 | 1 | 0 | 0 |
| sjlearmonth | 0 | 0 | 2 | 3 |
| Sinan81 | 0 | 0 | 1 | 4 |