Weekly GitHub Report for Tensorflow: January 27, 2025 - February 03, 2025
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 v2.18.0
1.2 Version Information:
The TensorFlow 2.18.0 release, created on October 21, 2024, introduces several key updates, including the addition of a fourth parameter to the TfLiteOperatorCreate function for a cleaner API, the disabling of TensorRT support in CUDA builds, and the implementation of hermetic CUDA support for more reproducible builds. Notably, TensorFlow now supports NumPy 2.0 by default, with changes in type promotion rules, and introduces enhancements in tf.lite such as support for TensorType_INT4 and TensorType_INT16 in various operations.
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.
-
Title: ValueError when adding TensorFlow Hub KerasLayer to Sequential model: This issue involves a ValueError encountered when adding a TensorFlow Hub KerasLayer to a Sequential model in TensorFlow version 2.17.1, despite the KerasLayer being a subclass of keras.Layer, which should theoretically be compatible according to the documentation. The user expected the KerasLayer to integrate seamlessly into the Sequential model, but the error suggests otherwise, indicating a potential documentation bug or version compatibility issue.
- The comments discuss a workaround involving the installation of tf-keras to resolve the error, but the user reports that this solution only partially works, as similar errors persist in other sections of the notebook. Further suggestions include modifying the import statements and using a different version of Keras, but the user continues to experience issues on their local machine despite these adjustments.
- Number of comments this week: 5
-
It doesn't support on python3.13: This issue is about the inability to install TensorFlow version 2.17 on Python 3.13 due to a lack of support, as indicated by the error message stating no matching distribution was found for TensorFlow. The problem arises because TensorFlow's release cycle does not align with Python's, leading to a delay in support for new Python versions, which has been a recurring issue since Python 3.8.
- The comments discuss the recurring issue of TensorFlow not supporting new Python versions immediately after their release, with users expressing frustration and suggesting workarounds like downgrading Python. Some comments highlight the complexity of TensorFlow's build system and dependencies, while others criticize the process as being inefficient and call for better synchronization with Python's release schedule. There is also a mention of potential technical challenges with Python 3.13's new features, and a suggestion that the issue is more about process management than technical barriers.
- Number of comments this week: 2
-
KeyError: "There is no item named 'PetImages\Cat\0.jpg' in the archive" When Running TensorFlow Locally(CPU) on Anaconda in VS Code.: This issue involves a KeyError encountered when running TensorFlow locally on a Windows machine using Anaconda in VS Code, specifically during the dataset download process, which works without issues on Google Colab. The error suggests that a specific file, 'PetImages/Cat/0.jpg', is missing from the archive, despite attempts to resolve the problem by updating software versions and verifying the dataset's contents.
- The comments discuss the need for documentation to reproduce the issue, with the user following a TensorFlow tutorial. Another user replicates the issue on Windows and suggests it may be due to file path parsing differences between operating systems, indicating a fix is underway. A link to a temporary workaround is provided, and a subsequent comment confirms that a fix has been merged into the TensorFlow Datasets repository, pending a new release.
- Number of comments this week: 2
-
How can I compile TensorFlowLite for Swift without Bitcode?": This issue is about a user seeking guidance on how to compile TensorFlowLite for Swift without Bitcode, as Apple has discontinued its use, and they are currently using version 2.17.0 from CocoaPods, which includes Bitcode. Additionally, the user inquires if there is a version of TensorFlowLite Swift that does not require Rosetta to run on ARM architectures.
- The comments provide a solution to disable Bitcode using a specific build command and suggest checking project settings in Xcode. The user confirms their Xcode version and the TensorFlowLiteSwift version they are using, and they are advised to try a newer version and report any issues with error logs for further investigation.
- Number of comments this week: 2
-
Tutorial "Multi-worker training with Keras" fails to complete: This issue reports a bug in the "Multi-worker training with Keras" tutorial, where the process fails when starting the second worker, resulting in errors related to CUDA and TensorFlow's inability to convert certain values to tensors. The user has confirmed the bug persists with TensorFlow Nightly and has provided detailed logs and system information to aid in troubleshooting.
- The comments discuss the need for additional details to troubleshoot the issue, with a request for more information about TensorFlow and library versions, and the user clarifies that they are using the tutorial's Jupyter notebook without custom code.
- Number of comments this week: 2
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.
As of our latest update, there are no stale issues for the project this week.
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 Version Discrepancies: Several issues have been reported regarding inconsistencies and bugs in TensorFlow version 2.18. These include inconsistent results between CPU and GPU executions for operations like
tf.image.adjust_hue,tf.raw_ops.BatchMatMulV2, andtf.raw_ops.BiasAddGrad, as well as a problem with thebatch_input_shapeargument for stateful LSTM models. These discrepancies suggest potential bugs or areas where documentation may need clarification.
- TensorFlow Hub KerasLayer Integration: A ValueError is encountered when adding a TensorFlow Hub KerasLayer to a
tf.keras.Sequentialmodel in TensorFlow version 2.17.1. The error message incorrectly states that only instances ofkeras.Layercan be added, despitehub.KerasLayerbeing a subclass ofkeras.Layer, indicating a potential bug or documentation issue.
- TensorFlow Lite and GPU Delegate on macOS: Users face challenges when building TensorFlow Lite with GPU delegate support on macOS platforms. Issues include buffer allocation errors with the OpenCL backend and C++ version compatibility and linking issues on a MacBook Air M2, raising questions about the feasibility of using a Mac for this development.
- Static C Libraries for Xtensa DSP Processors: There is a request for guidance on building static C libraries for Xtensa DSP processors using TensorFlow 2.8 on Ubuntu 20.04 with GCC 9.4.0. The user encounters problems with dangerous relocation and undefined references during compilation, indicating a need for more detailed documentation or support.
- Dynamic Range Quantization Clarification: Clarification is sought on the operations affected by Dynamic Range Quantization when using a specific post-training quantization tool. The inquiry focuses on whether only fully connected layers in transformers are impacted and if computations will be performed with int8 for both weights and activations.
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: 15
Summarized Issues:
- TensorFlow and CUDA Compatibility Issues: Users face challenges with TensorFlow not recognizing CUDA drivers on WSL setups, requiring downgrades to unsupported TensorFlow versions. These issues highlight the need for better compatibility and support for CUDA in TensorFlow environments on WSL.
- TensorFlow Lite Runtime Errors: Several users encounter runtime errors when using TensorFlow Lite models, such as YOLO11n-obb on Android and Super Resolution models, due to issues with GPU delegation and input-output mismatches. These errors prevent successful model deployment and require debugging of the conversion and inference processes.
- TensorFlow and Keras Compatibility Bugs: A bug in TensorFlow's autodiff function causes crashes due to incorrect attribute references in Keras layers, persisting across different Keras versions. This indicates a need for improved compatibility between TensorFlow and Keras.
- TensorFlow Lite Build and Compilation Issues: Users report problems building TensorFlow Lite for iOS and cross-compiling to aarch64, facing undefined symbols and exec format errors. These issues suggest a need for better support and documentation for building TensorFlow Lite across different platforms.
- TensorFlow GPU and XLA Compatibility Issues: Bugs in TensorFlow versions 2.17.0 and 2.18.0 cause errors with GPU operations and XLA_GPU_JIT devices, such as internal errors with
tf.math.floormodand missing OpKernel registrations. These issues highlight the need for robust GPU and XLA support in TensorFlow.
- TensorFlow C and C++ API Limitations: Users express concerns over the limited documentation and gradient support in TensorFlow's C and C++ APIs, seeking more comprehensive resources for effectively using these APIs. This indicates a need for improved documentation and support for TensorFlow's lower-level APIs.
- TensorFlow Import and Installation Errors: Users encounter ImportErrors and DLL load failures on Windows 11, despite multiple reinstallation attempts, due to potential issues with MSVC redistributables, CPU compatibility, or library mismatches. These errors suggest a need for clearer installation guidelines and compatibility checks.
- TensorFlow Datasets Library Bug: A bug in the TensorFlow Datasets library causes the
cats_vs_dogs.pyfile to fail in loading images due to improper handling ofZipFileobjects. This issue was resolved by modifying the function to avoid using multipleZipFileobjects, indicating a need for more robust dataset handling.
- TensorRT Inference Issues on Jetson AGX Xavier: Users report identical inference results from distinct models using TensorRT on Jetson AGX Xavier, suggesting potential issues with model conversion or inference processes. This highlights the need for reliable model conversion and inference support in TensorRT.
- TensorFlow Feature Request for Custom Loss Functions: A feature request highlights the inability to pass sample weights to custom loss functions in TensorFlow, resulting in errors, and seeks a solution similar to PyTorch's handling. This indicates a demand for more flexible loss function support in TensorFlow.
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. All other pull requests are grouped based on similar characteristics for easier analysis.
Pull Requests Opened This Week: 5
Key Open Pull Requests
1. [mlir][tosa] Update Tensorflow to match TOSA v1.0 specification (part 2): This pull request aims to update TensorFlow's TOSA (Tensor Operator Set Architecture) implementation to align with the TOSA v1.0 specification by adding NaN propagation mode support, changing the shift of the MUL operation to a tensor type, and modifying the start and size of the slice operation to use the TOSA shape type.
- URL: pull/86307
- Merged: No
2. [ROCM][NFC] TF-side adaptions required for BlasLt interface refactoring in XLA: This pull request involves making TensorFlow-side adaptations necessary for the refactoring of the BlasLt interface in XLA, and it is intended to be merged in conjunction with a related pull request in the OpenXLA repository.
- URL: pull/85835
- Merged: No
3. [oneDNN][CPU] fuse a matmul pattern: This pull request aims to optimize CPU performance by fusing a specific sub-graph pattern involving MatMul, BiasAdd, Mul, Add, and Elu operations into a more efficient _FusedMatMul operation using oneDNN, resulting in up to an 18% performance gain.
- URL: pull/86172
- Merged: No
- Associated Commits: 949cc
Other Open Pull Requests
- Max_pool1D API Example Addition: This pull request proposes the addition of an example for the Max_pool1D API in the TensorFlow project. The example aims to enhance the documentation by providing a practical use case for users to better understand the API's functionality.
- Segmentation Fault Fix in RaggedTensorToTensor: This pull request addresses a segmentation fault issue in the
RaggedTensorToTensoroperation of TensorFlow. It implements checks for empty values with non-empty row splits, ensuring consistent dimension size handling, and provides clear error messages for invalid input configurations.
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. All other pull requests are grouped based on similar characteristics for easier analysis.
Pull Requests Closed This Week: 15
Key Closed Pull Requests
1. [oneDNN] upgrading oneDNN version to 3.6.2: This pull request aims to upgrade the oneDNN library from version 3.5 to 3.6.2 in the TensorFlow project, incorporating several bug fixes and ensuring compatibility across various platforms such as Cascade Lake, Sapphire Rapids, and Granite Rapids, as evidenced by multiple commits addressing build updates and suggested changes.
- URL: pull/77927
- Merged: No
2. Fix typo in code logic of Conv3DTranspose(): This pull request addresses a typo in the code logic of the Conv3DTranspose() function within the TensorFlow project, specifically correcting the padding calculation to properly use the depth dimension's padding values instead of the height dimension's, and includes additional commits for adding a regression test along the depth dimension, removing test suite instantiation duplication, and adjusting the placement of new test cases for syntactical correctness.
- URL: pull/77968
- Merged: No
3. [mlir][tosa] Update Tensorflow to match TOSA v1.0 specification: This pull request updates TensorFlow's TOSA (Tensor Operator Set Architecture) implementation to align with the TOSA v1.0 specification by incorporating changes from several LLVM patches, including updates to convolution operators' accumulation types, Tile and Pad operations, and ensuring consistent operand ranks for element-wise operations.
- URL: pull/85608
- Merged: Yes
Other Closed Pull Requests
- TensorFlow Lite Interpreter
num_threadsparameter validation: This pull request modifies the validation of thenum_threadsparameter to align with documented behavior. It allows-1for implementation-defined threading and treats0as disabling multithreading, resolving a previous discrepancy.
- Spam reduction in issue creation: A pull request was merged to reduce spam from bot accounts by enforcing adherence to a specified template for issue creation. This change aims to improve the quality of issues submitted to the TensorFlow repository.
- Regular expression pattern correction in CMakeLists.txt: This pull request modifies the regular expression pattern in the CMakeLists.txt file to ensure strict filename matching. It prevents unintended directory name matches that were causing compilation and linking problems.
- Crash fix in
MapUnstagefunction: A pull request was successfully merged to address a crash in theMapUnstagefunction when thekeyis not a scalar. This fix resolves the issue reported in issue #72295.
- Documentation and typographical corrections: Several pull requests aimed at correcting typographical errors and broken links in the TensorFlow documentation. While some were not merged, one successfully updated broken links in the
index.mdfile.
- Unmerged pull requests: Various pull requests, including those for elementwise checks, a new operation proposal, and updates to
xla_ops.cc, were not merged into the TensorFlow project. These include attempts to resolve specific issues and add new features.
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 |
|---|---|---|---|---|
| mihaimaruseac | 9 | 1 | 0 | 80 |
| Venkat6871 | 3 | 3 | 0 | 29 |
| gaikwadrahul8 | 2 | 2 | 0 | 26 |
| tilakrayal | 3 | 2 | 0 | 21 |
| weilhuan-quic | 11 | 3 | 0 | 0 |
| arzoo0511 | 0 | 0 | 0 | 14 |
| LongZE666 | 0 | 0 | 12 | 1 |
| codinglover222 | 4 | 2 | 2 | 4 |
| pkgoogle | 0 | 0 | 0 | 12 |
| c8ef | 6 | 1 | 0 | 4 |