Weekly GitHub Report for Tensorflow - 2024-12-30 12:00:02
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.
-
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 arises because TensorFlow's release cycle does not align with the release of new Python versions, leading to a delay in support for the latest Python releases.
- The comments discuss the recurring issue of TensorFlow's delayed support for new Python versions, with users expressing frustration over the lack of compatibility with Python 3.13, especially since it is the default version in major distributions like Fedora 41. Some users suggest downgrading Python as a temporary solution, while others argue that the problem is more about process inefficiencies within the TensorFlow team rather than technical barriers. There are also discussions about the complexities of TensorFlow's build system and dependencies, which contribute to the delay in supporting new Python versions.
- Number of comments this week: None
-
TF - Problems when trying to use GPU on M3 Max: This issue involves a bug with TensorFlow 2.16.2 on an Apple M3 Max GPU, where the program crashes immediately when attempting to use the GPU, despite functioning correctly with the CPU. The problem appears to be related to memory allocation issues with TensorFlow's Metal backend, as indicated by the error message about a pointer being freed that was not allocated.
- The comments discuss troubleshooting steps, including verifying the Metal backend installation, using the CPU instead of the GPU, and ensuring the latest macOS and TensorFlow versions. A user confirms the code works on their M3 Max, suggesting potential issues with the user's Python version or Anaconda environment. Another user reports a similar issue on an M1 Air, and the original poster resolves the problem by switching from Anaconda to Brew. Further suggestions include trying a simplified Keras model, disabling GPU acceleration, and checking TensorFlow's memory management settings.
- Number of comments this week: None
-
DLL load failure: This issue is about a bug encountered when trying to run TensorFlow version 2.18 on Windows 10 with Python 3.12, where the user experiences a DLL load failure related to the
_pywrap_tensorflow_internalmodule. The error message suggests that the native TensorFlow runtime failed to load, and the user is seeking a solution to successfully execute their code.
- Several users report experiencing the same error, prompting a request for more information about the TensorFlow version and potential compatibility issues. Suggestions include checking for the MSVC 2019 redistributable, CPU support for AVX2 instructions, and ensuring the correct library installations. A user clarifies the TensorFlow version as 2.18.0, and another user is advised to open a separate issue for better tracking and assistance.
- Number of comments this week: None
-
tf.keras.Sequentialnot atf.Modulesince 2.16?: This issue is about a change in TensorFlow version 2.16, where thetf.keras.Sequentialmodel is no longer recognized as atf.Module, affecting the.variablesattribute in custom modules. The user is seeking clarification on whether this is an intended change and how to properly use atf.keras.Modelwithin atf.Modulein the updated version.- The comments discuss attempts to reproduce the issue on different TensorFlow versions, confirming the problem persists. A solution is provided, indicating that the issue arises from the default use of Keras 3 in TensorFlow 2.16 and later. The suggested fix is to install and use Keras 2, which resolves the issue. The original poster confirms the solution works but expresses disappointment over the lack of backward compatibility in TensorFlow updates.
- Number of comments this week: None
-
Cannot Convert 51 to a shape - Movenet pose classification tutorial: This issue involves a problem encountered in the Movenet pose classification tutorial, where the user is unable to convert the input shape '51' to a valid shape using TensorFlow 2.17.0 on a Windows 11 platform, both locally and in a Colab notebook. The user is seeking advice or insight into resolving the ValueError that arises when defining the model with
tf.keras.Input(shape=(51)).- The comments reveal that another user initially did not encounter the same error and provided a link to a working example, but the original poster clarified that their issue was specifically with building the model embedding. Further discussion indicated that the problem might be related to a temporary server issue affecting model downloads, and a workaround was suggested by manually downloading the model from Kaggle. The conversation continued with attempts to replicate the issue, and the original poster expressed a need for further investigation and updates on the problem.
- Number of comments this week: None
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:
- Build and Compilation Issues: Users are encountering build failures when compiling TensorFlow with GPU support on Windows due to missing header files and other dependencies. These issues often involve modifying build files like Bazel to resolve missing components. Despite these efforts, users continue to face challenges in successfully compiling TensorFlow with the necessary GPU support.
- Runtime and Import Errors: Several users report failures related to loading the native TensorFlow runtime, often due to missing or incompatible DLLs. These issues result in ImportErrors when attempting to import TensorFlow modules, and persist even with newer versions of TensorFlow. Users are seeking solutions to resolve these persistent runtime errors.
- GPU Compatibility and Performance: Users are facing difficulties with GPU compatibility and performance when using TensorFlow on various systems. Issues include lack of support for specific GPUs, inability to perform GPU computations, and failures in creating GPU plans for certain operations. These problems highlight the challenges in ensuring TensorFlow's GPU support across different hardware configurations.
- Model Conversion Challenges: Users are experiencing issues when converting models to TensorFlow Lite, particularly with integer quantization. These challenges include incorrect predictions and conversion failures, often accompanied by errors related to compatibility and data types. Users are exploring workarounds such as downgrading TensorFlow versions to address these conversion issues.
- General Frustration and Bug Reports: Some users express frustration with TensorFlow's design and functionality, submitting bug reports with minimal information. These reports often include sarcastic comments and lack detailed descriptions, making it challenging for others to provide assistance. Commenters frequently request more information to better understand and address the reported issues.
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: 27
Summarized Issues:
- TensorFlow Function Parameter Bugs: Several issues have been reported regarding bugs in TensorFlow functions when specific parameters are set to extreme values. For instance, the
tf.nn.max_pool3dfunction crashes with a floating point exception whenksizeis too large, andtf.raw_ops.AudioSpectrogramfails with segmentation faults or RAM errors with inappropriatewindow_sizevalues. Similarly,tf.raw_ops.RFFT3Dandtf.signal.rfft3dencounter overflow and assertion errors with large or zerofft_lengthvalues.
- TensorFlow 2.15 Specific Bugs: TensorFlow version 2.15 has been identified with multiple bugs affecting various operations. The
tf.unstackfunction fails withuint64data types, andtf.raw_ops.TensorScatterAddcrashes with high-rankindices. Additionally,tf.raw_ops.Batchandtf.raw_ops.TensorListSplitencounter "Check failed" errors with large parameter values.
- TensorFlow Assertion and Overflow Errors: Assertion failures and overflow errors are prevalent in TensorFlow functions when parameters are set to zero or excessively large values. The
tf.raw_ops.RFFT2Dandtf.signal.rfft2dfunctions fail with zerofft_length, andtf.signal.inverse_stftencounters assertion failures with largeframe_length. Thetf.compat.v1.nn.ctc_lossfunctions also face overflow errors with largelabels.
- TensorFlow Model and Layer Issues: Users have reported issues related to model execution and layer initialization in TensorFlow. A shape mismatch error occurs in version 2.17.0, and a
ValueErrorarises from improper initialization of theTimeDistributedlayer. Additionally, discrepancies in model predictions between different devices and performance issues in TensorBoard's Trace Viewer have been noted.
- TensorFlow Lite and Android Issues: Several issues have been reported regarding TensorFlow Lite and its integration with Android. Users face challenges in building TensorFlow Lite libraries for Windows x86 and preventing layer fusion during model conversion. Additionally, Android applications encounter tensor allocation failures and linking errors with the GPU delegate.
- TensorFlow Compilation and Import Errors: Compilation and import errors have been encountered in TensorFlow, particularly on Windows systems. A compile error occurs in
TrieRawHashMap.cppwhen building from source, and an "ImportError" prevents TensorFlow from starting on Windows 11 Pro. These issues highlight challenges in compatibility and setup on specific platforms.
- TensorFlow Performance and Optimization Issues: Performance discrepancies and optimization challenges have been reported in TensorFlow. The
keras.layers.GRUfunction behaves differently on GPU versus CPU, and enabling JIT compilation causes NaN values in the Dirichlet noise function. These issues require adjustments in function parameters or settings to achieve consistent performance.