Weekly GitHub Report for Tensorflow - 2025-01-06 12:00:37
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 Other Noteworthy Updates:
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.
- Failing to convert MobileNetV3Large to TFLite w/ Integer q: This issue involves the failure to convert a MobileNetV3Large model to TensorFlow Lite (TFLite) with integer quantization, resulting in incorrect predictions on Windows 10 and conversion errors on Windows Subsystem for Linux (WSL). The user reports that the model produces unrelated outputs when converted with TensorFlow 2.10 on Windows 10, and encounters an LLVM error during conversion on WSL with TensorFlow 2.16.
- The comments discuss various attempts to resolve the issue, including downgrading TensorFlow to version 2.14.1, which reportedly resolves the problem for some users. Another user suggests using the latest Keras version, which improves results but introduces a new issue when using a representative dataset, leading to worse outcomes. The discussion highlights compatibility issues between TensorFlow and Keras versions, with suggestions to try different TensorFlow versions and requests for more detailed code examples to further investigate the problem.
- Number of comments this week: None
-
It doesn't support on python3.13: This issue is about the inability to install TensorFlow version 2.17 on Python 3.13, as the installation process fails to find a compatible version of TensorFlow for this Python version. The problem is particularly significant because Python 3.13 is the default version in the latest Fedora 41 release, and users are unable to install TensorFlow on systems running this version of Python.
- The comments discuss the recurring issue of TensorFlow's delayed support for new Python versions, with users expressing frustration over the lack of support for Python 3.13. Some users suggest downgrading Python, while others highlight the importance of supporting the latest Python version due to its adoption in major distributions like Fedora. The conversation also touches on the complexities of TensorFlow's build process and dependencies, with some arguing that the lack of support is more of a process issue than a technical one. There is a call for better synchronization with Python's release schedule, and a tentative timeline for support in the next TensorFlow release is mentioned.
- Number of comments this week: None
-
tensorflow-opt-cuda: error running on Linux via GPU: This issue involves a bug where a TensorFlow program fails to run on a GPU using Manjaro Linux, despite working on a CPU and on Windows, with the error indicating a problem with the CUDA environment and potential library version mismatches. The user reports that while PyTorch runs fine on the GPU, TensorFlow crashes with an exit code 134, suggesting a segmentation fault or memory corruption issue.
- The comments discuss troubleshooting steps, including verifying GPU detection by TensorFlow, checking for CUDA and cuDNN version mismatches, and ensuring TensorFlow compatibility with installed CUDA versions. Suggestions include downgrading CUDA, using TensorFlow nightly builds, and enabling GPU memory growth. There is also mention of a persistent CUDA plugin registration error affecting many users, with some users suggesting waiting for a fix from the TensorFlow team.
- Number of comments this week: None
-
Mixing Keras Layers and TF modules.: This issue involves a user experiencing difficulties when mixing Keras layers with TensorFlow modules, specifically noting that
tf.Modulecan tracetf.Variablebut not variables fromtf.kerasortf.keras.Variable. The user is seeking guidance on how to resolve tracing issues when using Keras layers withintf.Module, as they are encountering problems with the current TensorFlow and Python versions, and are unsure of the supported path forward.- The comments discuss a change in TensorFlow 2.16 where Keras layers no longer extend
tf.Module, causing issues with variable tracking. It is suggested to raise the issue in the Keras repository as it relates more to Keras. The user expresses concern about missing features in Keras and questions if TensorFlow will provide replacements for common layers, but it is noted that such replacements are unlikely. - Number of comments this week: None
- The comments discuss a change in TensorFlow 2.16 where Keras layers no longer extend
-
Tensorflow BackupAndRestore method does not work: This issue is about a bug in TensorFlow version 2.17 where the
BackupAndRestoremethod does not function as expected because the model is not built before calling thefit()method, resulting in aValueError. The problem arises because the model needs to be explicitly built or defined with an input shape before using theBackupAndRestorecallback to ensure proper backup and restoration during training.- The comments discuss the requirement for the model to be built before using the
BackupAndRestorecallback, offering solutions such as defining the input shape or usingmodel.build(). A pull request addressing the issue has been merged, and users are asked to verify if the problem persists. There are additional queries about the applicability of the solutions when using a normalization layer as the input layer, with some users encountering warnings and errors when attempting the proposed solutions. - Number of comments this week: None
- The comments discuss the requirement for the model to be built before using the
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: 10
Summarized Issues:
- TensorFlow Model Summary Bug: The
model.summary()function in TensorFlow is not displaying the correct output shape for custom layers, showingNoneinstead of the expected shape. This issue suggests that the problem might be resolved by reshaping the output within the custom layer'scallmethod. The incorrect display of output shapes can lead to confusion and misinterpretation of model architecture.
- Cross-Compilation and Build Errors: Users are facing issues with cross-compiling TensorFlow Lite for RISC systems on x86_64 machines due to undefined '@llvm_toolchain' repositories. Despite modifying the WORKSPACE file to point to a local LLVM installation, the Bazel build command fails. These errors hinder the development process and require a clear definition of the toolchain in the build configuration.
- TensorFlow and Keras Integration Issues: Changes in Keras 3.0, which no longer extends
tf.Module, are causing issues with TensorFlow'stf.Moduletracing variables fromtf.keraslayers. This change complicates the integration of Keras layers within TensorFlow modules. Users are seeking alternative solutions or raising the issue with the Keras team to address these integration challenges.
- TensorFlow Backup and Restore Bug: The
BackupAndRestoremethod in TensorFlow is encountering a bug where the example code fails due to the model not being built before callingfit(). This results in aValueError, which can be resolved by explicitly defining the input shape or building the model beforehand. Such issues can disrupt the workflow and require users to adjust their model setup process.
- TFLITE NMS Kernel Output Inconsistency: There is a bug in the TFLITE NMS kernel where its output is inconsistent with TensorFlow NMS, appending zeros to the "selected_indices" output up to "max_output_size". This inconsistency leads to slow computation and out-of-memory errors on Android devices. A fix is requested to ensure identical outputs between TFLITE and TensorFlow NMS.
- TensorFlow Source Code Compatibility Issues: The TensorFlow 2.4.1 source code has a bug where the
label_image.pytest case fails due to a missing 'GraphDef' attribute. This requires importing fromcompat.v1for backward compatibility with TensorFlow 1.x. Despite attempts to resolve it, additional errors related to missing files persist, complicating the testing process.
- Performance and Initialization Issues on Android: Users are experiencing performance lag and initialization failures when using the GPU delegate for object detection and image classification in Android Kotlin projects. These issues affect the functionality and efficiency of applications. Users are seeking advice to resolve these problems and improve performance.
- TensorFlow Dataset Handling and Compatibility Issues: A KeyError occurs when running TensorFlow locally on Anaconda in VS Code, indicating a missing file in the dataset archive. The same code executes successfully on Google Colab, suggesting potential differences in dataset handling or compatibility issues between local and cloud environments. This discrepancy raises questions about the consistency of dataset management across platforms.
- TensorFlow Import Error on Windows 11: Users are encountering a TensorFlow import error on Windows 11 after reinstalling Python and their Anaconda environment. The error message indicates a failure to load the native TensorFlow runtime due to a DLL initialization problem. Despite trying various Python versions and configurations, the issue persists, hindering the use of TensorFlow on Windows 11.
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: 19
Summarized Issues:
- TensorFlow Import Errors on Windows: Users have reported ImportErrors when attempting to load the native TensorFlow runtime on Windows systems, often due to DLL load failures. These issues are frequently linked to outdated CPU architectures or missing dependencies, such as the MSVC 2019 redistributable. The errors are consistent across different versions of Windows and Python, indicating a broader compatibility problem.
- TensorFlow Bugs and Crashes: Several bugs have been identified in TensorFlow, causing crashes and unexpected behavior. These include a "Check failed" error in
tf.strings.unsorted_segment_joinand integer overflow intf.raw_ops.DenseBincount. Additionally, issues with GPU support on Apple M3 Max and Android have been reported, highlighting compatibility challenges.
- TensorFlow Lite and Model Conversion Issues: Users have encountered problems when working with TensorFlow Lite, particularly during model conversion and deployment. These issues include exceptions during conversion from .h5 to .tflite and discrepancies in model behavior between Python and Android. Such problems often stem from input tensor order and data type mismatches.
- TensorFlow Version Compatibility and Installation Issues: Compatibility issues have arisen with TensorFlow installations, particularly concerning Python version requirements. Users have reported failures when attempting to install TensorFlow 2.x on Python 2.7.5, as the newer TensorFlow versions require Python 3.7 or above, leading to installation errors.
- TensorFlow Documentation and Naming Suggestions: There have been proposals to update TensorFlow's documentation and naming conventions. Suggestions include renaming "TensorFlow" to "Tensor Flow" for clarity and updating the CODE_OF_CONDUCT document to a newer version. These proposals aim to improve user experience and project governance.
- TensorFlow Deprecation Warnings and Compilation Issues: Users compiling TensorFlow with newer C++ standards have encountered deprecation warnings. Specifically, the use of
std::is_podin C++20 has been deprecated, prompting suggestions to replace it withis_standard_layout && is_trivialto avoid warnings and ensure compatibility.
- Unexpected Behavior in TensorFlow Operations: Users have reported unexpected behavior in TensorFlow operations, such as incorrect gradient computations and persistent warning messages. These issues disrupt expected outcomes and console readability, prompting users to seek clarification and solutions.
- Docker Image GPU Support Issues: The
tensorflow:2.18-gpu-jupyterDocker image has been reported to lack GPU support due to missing CUDA libraries. Users are required to manually installtensorflow[and-cuda]to enable GPU functionality, which contradicts previous documentation and user expectations.
- User Frustration with TensorFlow Design: A user expressed frustration with TensorFlow 2.16.1 on Linux Ubuntu 22.04, describing it as poorly designed. The issue provides minimal information for reproducing the problem, highlighting user dissatisfaction with the software's usability.
- TensorFlow Layer Compatibility Issues: A bug in TensorFlow 2.8 was reported where
tensorflow_hub.KerasLayerwas not recognized as an acceptable layer in a sequential model. This issue was resolved by adjusting import statements and using a compatible version of Keras, indicating the importance of version compatibility in TensorFlow projects.