Weekly GitHub Report for Tensorflow: August 18, 2025 - August 25, 2025 (12:03:59)

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
  • 1.2. Other Noteworthy Updates
• II. Issues
  • 2.1. Top 5 Active Issues
  • 2.2. Top 5 Stale Issues
  • 2.3. Open Issues
  • 2.4. Closed Issues
  • 2.5. Issue Discussion Insights

• III. Pull Requests
  • 3.1. Open Pull Requests
  • 3.2. Closed Pull Requests
  • 3.3. Pull Request Discussion Insights

• IV. Contributors
  • 4.1. Contributors

I. News

1.1 Recent Version Releases:

The current version of this repository is v2.19.0

1.2 Version Information:

Released on March 5, 2025, TensorFlow version 2.19.0 introduces breaking changes to the tf.lite API, including the deprecation of tf.lite.Interpreter in favor of ai_edge_litert.interpreter and changes to certain C++ constants for improved API flexibility. Key updates also include runtime support for the bfloat16 data type in the tfl.Cast operation, alongside the discontinuation of standalone libtensorflow package publishing, with the package still accessible 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.

1. ValueError: Shape mismatch in layer #1 (named stem_conv)for weight stem_conv/kernel. Weight expects shape (3, 3, 1, 48). Received saved weight with shape (3, 3, 3, 48): This issue reports a ValueError caused by a shape mismatch when loading pretrained weights for the EfficientNetB4 model in TensorFlow, where the model expects an input with one channel but receives weights for three channels. The user encounters this error while trying to instantiate the model with an input shape of (256, 256, 3) and pretrained ImageNet weights, and notes the same problem occurs with EfficientNetB5.

   • The comments include a request for the full reproducible code, which the user provides, clarifying the intent to load the pretrained model and highlighting the channel mismatch error. Another participant tests the code on different TensorFlow versions without reproducing the error and shares a Colab notebook for reference. The user acknowledges the assistance but remains uncertain about the root cause of the issue.
   • Number of comments this week: 4

2. CPU produces incorrect gradients / zero outputs for [OPERATION_NAME]: This issue reports a bug where running a specific TensorFlow operation on CPU produces incorrect gradients containing unexpected zeros, while the same operation on GPU yields correct gradient outputs. The problem appears to be environment-specific, occurring with Python 3.8 and TensorFlow 2.19 on Ubuntu 20.04, and does not reproduce on newer Python versions or in TensorFlow Nightly builds tested on Colab.

   • The commenters conducted thorough tests on multiple TensorFlow versions and Python environments, confirming the issue does not reproduce on newer Python versions (3.10, 3.12) or in TF Nightly on Colab. They recommended upgrading Python and TensorFlow versions and using isolated environments like Conda to avoid compatibility problems. The original reporter acknowledged the advice and plans to test in a fresh environment, appreciating the detailed guidance and support.
   • Number of comments this week: 4

3. Tensor Flow Lite INT8 Issue: This issue concerns difficulties encountered when converting a Keras .h5 model to TensorFlow Lite format with INT8 quantization, where the user observes that while float16 conversion results in some accuracy drop, the INT8 quantized model produces heavily biased predictions. The user has shared detailed code for the conversion and evaluation process and is seeking advice on whether the INT8 quantization might be degrading the model weights excessively and how to properly handle the conversion and inference steps.

   • The comments provide practical suggestions including removing manual setting of INT8 input/output types in the converter, clipping inputs during inference, applying sigmoid to outputs, and verifying that all operations are quantized correctly. Another commenter advises simplifying the inference code by letting the TFLite interpreter handle quantization internally and ensuring the representative dataset is properly normalized and sufficiently large to improve quantization accuracy, while also requesting a reproducible example if issues persist.
   • Number of comments this week: 3

4. Cannot call first version of tf.range() providing a dtype using positional arguments: This issue reports a bug in TensorFlow version 2.9.3 where the first version of the tf.range() function cannot accept a dtype argument using positional parameters, resulting in errors when attempting to specify the data type directly. The user demonstrates that calling tf.range(5, 2, tf.float32) or including a name argument causes type conversion errors, indicating that the function mistakenly interprets the dtype as the delta parameter.

   • The comments clarify that the function is interpreting the tf.float32 argument as the delta parameter rather than the data type, effectively selecting the second function signature. Suggestions include either providing a clearer error message or adding a default value for delta in the Python wrapper to handle this case more gracefully.
   • Number of comments this week: 3

5. ForwardAccumulator: Improve error message when primals and tangents shapes mismatch: This issue requests an improvement to the error message in TensorFlow's ForwardAccumulator when the shapes of primals and tangents do not match. Currently, a low-level reshape error is raised, which is confusing; the user suggests adding an early, clear shape validation that raises a descriptive ValueError during initialization to improve usability.

   • The discussion includes an apology for earlier noise and presents a pull request that adds early shape validation in the ForwardAccumulator constructor, replacing the obscure reshape error with a clear ValueError. The fix is well received and appreciated for making the API easier to use and the error messages more informative.
   • 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.

1. TF-TRT Warning: Could not find TensorRT: This issue describes a problem where the user is encountering a warning that TensorRT cannot be found when running TensorFlow on an Ubuntu 22.04 system with an RTX 3050 Ti GPU and CUDA 12.4. Despite using the NVIDIA driver version 535 (after reverting from 550) and reinstalling TensorFlow multiple times, the user is unable to resolve the TensorRT detection error and is seeking assistance to fix this configuration problem.
2. SystemError in tf.ensure_shape and tf.compat.v1.ensure_shape when dtype of shape is tf.uint64 and its value is too large.: This issue reports a bug in TensorFlow where using tf.ensure_shape or tf.compat.v1.ensure_shape with a shape tensor of type tf.uint64 containing very large values close to 2^64 causes a SystemError and OverflowError. Specifically, when such large values are passed in eager execution mode, the functions fail with an error related to the built-in isinstance function, indicating improper handling of large unsigned 64-bit integers in shape validation.
3. Feature Request: Integrate different Digital Signal Processing into tf.signal: This issue is a feature request proposing the integration of advanced digital signal processing (DSP) functionalities, similar to those found in the julius library, into TensorFlow's tf.signal module. The requester highlights the current lack of sophisticated audio data augmentation tools within TensorFlow compared to PyTorch and suggests that adding these capabilities would enhance audio model development by enabling native, efficient preprocessing and augmentation workflows.
4. [DOCS] Missing complex input for Round op: This issue reports a documentation bug in TensorFlow where the Round operation is described as supporting complex tensor inputs, but in practice, attempting to use a complex tensor with this operation results in an error, requiring users to round the real and imaginary parts separately. The user provides a reproducible example and error logs showing that the operation fails due to the absence of a suitable device kernel for complex inputs, highlighting a discrepancy between the official documentation and actual functionality.
5. tf.raw_ops.Unbatch aborts with "Check failed: d < dims()": This issue reports a bug in TensorFlow version 2.17 where the operation tf.raw_ops.Unbatch aborts with a fatal check failure error "Check failed: d < dims()" when executed, causing the program to crash. The problem occurs on Linux Ubuntu 20.04.3 LTS with Python 3.11.8 and has been reproduced using TensorFlow Nightly, with a minimal code snippet provided that triggers the abort due to an invalid tensor dimension check.

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: 13

Summarized Issues:

• Convolution and Model Weight Mismatches: Several issues report mismatches and errors related to convolution operations and model weight shapes in TensorFlow. These include discrepancies between INT8 convolution kernels causing output differences and shape mismatches when loading pretrained EfficientNetB4 weights, leading to failures in model loading.
• [issues/99072, issues/99122]

• Error Handling and User Messages: Multiple issues highlight the need for clearer, more user-friendly error messages in TensorFlow. Problems include confusing low-level reshape errors in ForwardAccumulator, unclear internal assertion errors from Autograph's conditional conversion, and requests for improved error messages when unsupported data types are used in operations like argsort.
• [issues/99160, issues/99162, issues/99164]

• TensorFlow Operation Bugs and Type Mismatches: There are bugs involving type and axis mismatches in TensorFlow operations, such as take_along_axis failing due to axis mismatch despite correct dtypes, and tf.range() misinterpreting dtype arguments causing type conversion errors.
• [issues/99163, issues/99230]

• Gradient Computation and Backend Discrepancies: An issue reports incorrect gradient outputs on CPU for a specific operation in TensorFlow 2.19 with Python 3.8, where gradients contain unexpected zeros, while the GPU backend produces correct results, indicating a backend-specific bug affecting gradient computation.
• [issues/99165]

• Build and Linking Failures: Several issues describe build and import failures of TensorFlow on specific platforms and configurations. These include missing symbols in pybind11_protobuf on ppc64le with Python 3.12, linking errors due to abseil version mismatches on Ubuntu 24.04 with CUDA 12.9.1, and permission errors accessing build logs in the official builds section.
• [issues/99246, issues/99255, issues/99273]

• Documentation and Parameter Clarifications: One issue requests updates to documentation and error messages for the round_mode parameter in tf.quantization.quantize(), clarifying that HALF_TO_EVEN rounding is only supported with SCALED mode and fixing a typographical error in the error message.
• [issues/99286]

• Contributor License Agreement (CLA) Access Issues: There is a reported problem where users in mainland China are unable to sign the TensorFlow Contributor License Agreement, blocking contributions from that region.
• [issues/99117]

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: 6

Summarized Issues:

• GPU support and hardware compatibility issues: Several issues report problems related to GPU support and hardware compatibility. One issue describes TensorFlow 2.19.0 not being GPU-enabled on an Alma Linux server with CUDA 12.3 and NVIDIA GH200 GPU, while another details a SIGILL error caused by disabled AVX2 support in VirtualBox when running TensorFlow C++ on an Ubuntu guest. These problems highlight challenges in ensuring TensorFlow runs correctly with specific hardware and virtualization environments.
• [issues/98115, issues/99118]

• Import errors and DLL load failures: Multiple issues involve ImportErrors due to DLL load failures or missing native dependencies. One issue involves TensorFlow 2.0 failing to import due to version and Python environment incompatibilities, while another describes a DLL initialization failure on Windows 10 with Python 3.12 caused by missing Microsoft Visual C++ Redistributable or import order conflicts. These errors emphasize the importance of compatible environments and dependencies for successful TensorFlow imports.
• [issues/99055, issues/99291]

• Precision and computation discrepancies: There is a reported precision bug where GPU computations of tf.math.segment_max and tf.math.segment_min on large floating-point tensors differ abnormally from CPU outputs. This issue points to potential inconsistencies in numerical results between different hardware backends in TensorFlow.
• [issues/97805]

• Spam and irrelevant content: One issue is identified as spam with no substantive content or relevance to TensorFlow, making it inappropriate for the issue tracker.
• [issues/99283]

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: 3

Key Open Pull Requests

1. forwardprop: raise clear shape-mismatch error in ForwardAccumulator ctor: This pull request adds early shape validation to the non-batched tf.autodiff.ForwardAccumulator constructor to ensure that primals and tangents have compatible shapes, raising a clear ValueError at construction time to prevent later opaque reshape errors.

• URL: pull/99195

• Merged: No

• Associated Commits: 36d3c, 6efea

2. Fix the Doc and Error Format of round_mode in func tf.quantization.quantize(): This pull request aims to correct the documentation and improve the error formatting related to the round_mode parameter in the tf.quantization.quantize() function.

• URL: pull/99287

• Merged: No

• Associated Commits: 3e31e, c9220

3. Align TFLite CMake XNNPACK pin to upstream commit used by Bazel: This pull request updates the TFLite CMake configuration to align the XNNPACK commit pin with the upstream commit used by Bazel, replacing a non-existent commit with the correct upstream revision to unblock builds with XNNPACK enabled and resolve related issues.

• URL: pull/99134

• Merged: No

• Associated Commits: c590e

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: 8

Key Closed Pull Requests

1. Fix crash in tf.audio.decode_wav with invalid negative parameters: This pull request addresses and resolves a crash in the TensorFlow function tf.audio.decode_wav caused by invalid negative parameters being passed to TensorShape, which expects non-negative dimensions.

• URL: pull/99057

• Merged: No

• Associated Commits: 27c8c, f3e41, fb5be, bb92f, 7fedb

2. GPU: Fix overflow inconsistency in cumsum on CPU/GPU: This pull request fixes a platform-dependent overflow inconsistency in the cumulative sum (cumsum) operation between CPU and GPU in TensorFlow Lite’s GPU delegate by implementing higher-precision accumulation and modifying GPU shader logic to ensure consistent and overflow-safe results.

• URL: pull/97693

• Merged: 2025-08-19T15:34:37Z

• Associated Commits: 858e9

3. Improve grammar in regularizer error message: This pull request improves the grammar in the error message related to invalid regularization penalty numbers by changing the phrase "is not a property value" to "is not a valid property value" for enhanced clarity and correctness.

• URL: pull/98645

• Merged: 2025-08-18T08:10:10Z

• Associated Commits: 9581e

Other Closed Pull Requests

• Bug Fixes in TensorFlow Functions: Multiple pull requests address critical bug fixes in TensorFlow functions to improve stability and correctness. One fixes a crash in tf.audio.decode_wav caused by invalid negative parameters passed to TensorShape, while another updates SparseApplyAdagrad CPU implementation to skip out-of-range indices, aligning behavior with the GPU version and removing error checks that caused discrepancies.
• [pull/99058, pull/99116]

• Documentation Improvements: Several pull requests focus on enhancing TensorFlow documentation for better clarity and accuracy. These include clarifying the optional dtype parameter in tf.constant() and fixing typos in docstrings to improve overall documentation quality.
• [pull/99092, pull/99207]

• Shape Validation Enhancements: A pull request introduces early shape validation in the non-batched tf.autodiff.ForwardAccumulator constructor to ensure compatible shapes for primals and tangents. This change raises a clear ValueError at construction time, preventing opaque reshape errors during forward propagation.
• [pull/99194]

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
tensorflower-gardener	184	0	0	0
mihaimaruseac	0	0	0	48
ezhulenev	35	0	0	0
Aaraviitkgp	16	5	0	10
djxnqp	0	0	31	0
springcrane	0	0	8	6
WillFroom	13	0	0	0
psamanoelton	2	2	0	9
beckerhe	12	0	0	0
wsmoses	12	0	0	0