Weekly GitHub Report for Tensorflow: September 15, 2025 - September 22, 2025 (12:07:02)

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
  • 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 LiteRT (tf.lite) API, including the deprecation of tf.lite.Interpreter in Python with a migration path to ai_edge_litert.interpreter, and updates to C++ constants for better API compatibility. Key improvements include runtime support for the bfloat16 data type in the tfl.Cast operation, alongside the discontinuation of separate libtensorflow package publishing, while still allowing unpacking from 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. .compile throws ValueError: Could not interpret optimizer identifier: <keras.src.optimizers.adam.Adam object at 0x7f419cfc5010>: This issue concerns a ValueError encountered when calling the .compile method on a TensorFlow Keras model, where the optimizer passed is an instance of the standalone Keras Adam optimizer rather than TensorFlow's Keras Adam optimizer. The error arises because the model, subclassed from tf.keras.Model by Hugging Face's TFDistilBertForTokenClassification, expects an optimizer from tensorflow.keras.optimizers, but the code uses the Adam optimizer from the separate standalone Keras package, causing a mismatch.

   • The comments discuss the root cause being the use of the standalone Keras Adam optimizer instead of TensorFlow's Keras Adam, with suggestions to import the optimizer from tensorflow.keras.optimizers to fix the issue. The original poster insists they are already doing so, but others point out that the traceback clearly shows the standalone Keras optimizer is being used, possibly due to environment conflicts causing the wrong package to be imported.
   • Number of comments this week: 5

2. Attention layer is not able to predict on Macos: This issue reports a bug where the Attention layer in TensorFlow fails to predict correctly on MacOS M1 and M4 devices using Python 3.12, resulting in broadcast errors due to incompatible shapes when masks are applied. The problem does not occur on Linux or on MacOS with Python 3.11 and TensorFlow 2.13.0, indicating a regression likely introduced around TensorFlow 2.16.0 and the Keras 2 to 3 upgrade.

   • The single comment identifies the root cause as a breaking change introduced in TensorFlow 2.16.0 coinciding with the upgrade from Keras 2 to 3, suggesting the issue is related to recent updates in these libraries.
   • Number of comments this week: 1

3. Segmentation fault in TensorFlow API tf.raw_ops.CropAndResize: This issue reports a segmentation fault occurring in TensorFlow version 2.21 when using the tf.raw_ops.CropAndResize API with extremely large values in the image tensor parameter, which causes the service to crash due to insufficient validation. The user provides a minimal reproducible example and detailed environment information, confirming the bug persists in the nightly build and requests assignment to fix the problem.

   • The single comment expresses interest in taking ownership of the issue to contribute a fix, indicating a willingness to address the bug.
   • Number of comments this week: 1

4. TF_CPP_MIN_LOG_LEVEL=3 doesn't silence import warnings in tensorflow 2.19: This issue reports that setting the TF_CPP_MIN_LOG_LEVEL environment variable to 3 in TensorFlow 2.19 does not suppress certain import-time warning messages related to CUDA and GPU initialization, which appear even on systems without a GPU. The user is seeking clarification on whether this behavior is expected, as these warnings remain visible despite the log level setting intended to silence them.

   • The comment confirms that this is expected behavior because these early GPU/CUDA warnings are emitted before TensorFlow’s logging system is initialized, so the environment variable cannot suppress them. The messages are harmless and only appear during startup, with no impact on normal execution or training, and the only way to hide them would be to redirect stderr or modify TensorFlow’s source code.
   • 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.

1. TF-TRT Warning: Could not find TensorRT: This issue describes a problem where TensorFlow on Ubuntu 22.04 cannot detect TensorRT, resulting in a warning message "TF-TRT Warning: Could not find TensorRT" despite having compatible NVIDIA drivers and CUDA versions installed. The user reports difficulties with driver compatibility on a multi-monitor setup using an RTX 3050 Ti GPU, and despite attempts to uninstall and reinstall TensorFlow, the TensorRT integration remains unresolved, impacting their ability to proceed with machine learning development.
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 comprehensive audio data augmentation tools within TensorFlow compared to PyTorch and suggests that adding these capabilities would enhance audio model training by enabling native, efficient preprocessing and augmentation workflows.
4. [DOCS] Missing complex input for Round op: This issue highlights 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 manually round the real and imaginary parts separately. The reporter reproduces the problem on TensorFlow version 2.15.0 running on MacOS Sonoma with Python 3.9, and provides error logs showing that the operation fails to find a suitable device kernel for complex inputs despite the documentation indicating such support.
5. tf.raw_ops.Unbatch aborts with "Check failed: d < dims()": This issue reports a bug in TensorFlow version 2.17 where the tf.raw_ops.Unbatch operation aborts with a fatal check failure error "Check failed: d < dims()". The problem occurs when running a specific code snippet on Linux Ubuntu 20.04.3 with Python 3.11.8, causing the program to crash due to an invalid dimension check in the tensor shape handling.

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: 14

Summarized Issues:

• TensorFlow compilation and optimizer compatibility issues: There is a bug where calling .compile on a Hugging Face TensorFlow model raises a ValueError because the optimizer passed is an instance of the standalone Keras Adam optimizer rather than TensorFlow Keras’s Adam, causing a mismatch. This prevents TensorFlow from recognizing the optimizer identifier and compiling the model correctly.
• issues/100330

• XLA JIT compilation incompatibilities: Enabling XLA JIT compilation with @tf.function(jit_compile=True) causes failures when compiling functions containing unsupported operations such as tf.print (due to StringFormat not supported on XLA_TPU_JIT), tf.nn.max_pool_with_argmax (unsupported MaxPoolWithArgmax), and tf.image.crop_and_resize (unsupported CropAndResize). These unsupported operations result in compilation errors and prevent the use of XLA JIT for these functions.
• issues/100517, issues/100519, issues/100521

• TensorFlow crashes and validation errors with extreme or invalid input values: TensorFlow 2.21 crashes due to insufficient validation in APIs such as tf.nn.separable_conv2d when given extremely large or negative strides values, and tf.raw_ops.CropAndResize when provided with extremely large values in the image tensor parameter. These issues cause the TensorFlow service to abort or segfault due to invalid input handling.
• issues/100374, issues/100375

• Platform-specific runtime errors and crashes: TensorFlow crashes with a std::system_error on machines with advanced AVX-512 CPU flags, while working fine on machines with only AVX and AVX2. Additionally, the TensorFlow Attention layer fails to predict on MacOS M1 and M4 devices using Python 3.12 due to broadcast shape errors with input masks, despite working on Linux and MacOS with Python 3.11.
• issues/100335, issues/100332

• TensorFlow Lite build and deployment issues: The TensorFlow Lite CMake installation fails because the export target "tensorflow-liteTargets" depends on targets like "pthreadpool", "xnnpack-delegate", and "XNNPACK" that are not included in any export set, causing build failures when creating installable packages. Also, TensorFlow Lite Task Vision and GPU Delegate Plugin libraries for Android lack required 16KB page alignment, causing app crashes and blocking updates on newer Android devices.
• issues/100686, issues/100533

• Installation and environment compatibility problems: Installing the latest tensorflow[cuda] package on Windows 11 with Python 3.13 fails due to the absence of a compatible nvidia-nccl-cu12 dependency version, causing installation errors in virtual environments. Additionally, setting TF_CPP_MIN_LOG_LEVEL to silence warnings in TensorFlow 2.19 does not suppress early GPU/CUDA-related warnings that appear before logging is initialized, resulting in persistent warning messages.
• issues/100689, issues/100442

• TensorFlow numerical operation bugs on specific platforms: On Windows 10 with TensorFlow 2.19, functions like tf.clip_by_value, tf.minimum, and tf.maximum produce incorrect results when applied to int64 tensors with int64 bounds, failing to properly clip values outside the specified range. This issue does not occur on Linux, indicating a platform-specific numerical bug.
• issues/100590

• Build support for new development tools: There is a feature request to add support for building TensorFlow on iOS using Xcode 26, as the current build fails with Xcode 26 but succeeds with Xcode 16.2, indicating a need to update build configurations for newer toolchains.
• issues/100434

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: 4

Summarized Issues:

• Documentation Clarifications on Function Usage: Several issues highlight the need for clearer documentation regarding function parameters and their constraints in TensorFlow. These include specifying incompatible formatting options in tf.strings.as_string(), the requirement of providing pad_values when padding_width is set in tf.strings.ngrams(), and the valid naming rules for the container argument in tf.raw_ops.ExperimentalThreadPoolHandle. Clear documentation is necessary to prevent user confusion and errors when using these functions.
• issues/100497, issues/100500, issues/100522

• User Behavior and Project Impact: One issue reports a user admitting to spamming the project due to lack of productive activities. This highlights concerns about user conduct and its potential negative impact on the project environment.
• issues/100573

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. fix(saved_model_cli): reject empty feature lists with clear ValueError: This pull request introduces early input validation in the _create_example_string function of saved_model_cli to reject empty feature lists by raising a clear ValueError, preventing a late IndexError during TFExample building without affecting valid inputs.

• URL: pull/100567

• Merged: No

• Associated Commits: a0f5f, 935ce

2. Fix out of bounds error in LstmStepInteger8x8_8 output gate: This pull request addresses an out-of-bounds error in the output gate recurrent part of the LstmStepInteger8x8_8 function by correcting the incorrect indices used for the intermediate_scale_a, intermediate_scale_b, and intermediate_zp arrays to ensure consistency with other gates and prevent accessing elements beyond the array size.

• URL: pull/100346

• Merged: No

• Associated Commits: 67c22

3. Add ksize validation to QuantizedMaxPool: This pull request adds validation to the ksize parameter in the QuantizedMaxPool MKL kernel by ensuring that each dimension of the input tensor is greater than or equal to the corresponding ksize value, thereby preventing invalid arguments that could cause runtime errors.

• URL: pull/100381

• Merged: No

• Associated Commits: b4916

Other Open Pull Requests

• Post LLVM-upstream fixes: This pull request addresses fixes related to the post llvm-upstream changes referenced by issue #152856 in the TensorFlow project. It ensures compatibility and stability following the upstream updates.
• pull/100456

• Input validation and unit tests for tf.nn.separable_conv2d strides: This pull request adds unit tests for the tf.nn.separable_conv2d function to validate stride arguments by rejecting negative, excessively large, and non-integer values. These changes prevent confusing runtime failures and improve input validation consistency.
• pull/100506

• SavedModel CLI input sanitization and validation: This pull request improves the robustness of the SavedModel CLI by sanitizing tag_set and --op_denylist inputs to trim whitespace and remove empty items. It also introduces stricter validation that raises clear errors on whitespace-only inputs and adds unit tests to ensure correct handling.
• pull/100570

• TensorFlow shared library compilation and symbol versioning: This pull request proposes compiling the TensorFlow shared library with the -Bsymbolic-functions flag and properly versioning symbols in tensorflow_framework. These changes address issue #99991 to improve symbol management.
• pull/100606

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: 5

Key Closed Pull Requests

1. Fix error handling in CropAndResizeOp: replace absl::InvalidArgumentError with errors::InvalidArgument: This pull request addresses incorrect error handling in the CropAndResize kernel by replacing absl::InvalidArgumentError with TensorFlow's standard errors::InvalidArgument to prevent segmentation faults caused by mixing different error types and to ensure consistent and clear error propagation when invalid inputs are provided.

• URL: pull/100400

• Merged: No

• Associated Commits: 8dba3

2. I'm a spammer and I spam: This pull request, titled "I'm a spammer and I spam," appears to be an unmerged contribution to the TensorFlow project containing a commit with the message "Create Ten50rflowso0" and includes optional extended descriptions.

• URL: pull/100487

• Merged: No

• Associated Commits: a4880

3. Clarify the limitation between scientific and shortest in func tf.strings.as_string(): This pull request clarifies the limitation between the scientific and shortest formatting options in the function tf.strings.as_string() to improve understanding and usage.

• URL: pull/100498

• Merged: Yes

• Associated Commits: b9587

Other Closed Pull Requests

• Error message clarification in tf.strings.ngrams(): This pull request improves the usability of the tf.strings.ngrams() function by clarifying the error message raised between the pad_values and padding_width parameters. It directly addresses issue #100500 to help users better understand parameter conflicts.
• pull/100501

• Fixes in tf.raw_ops.ExperimentalThreadPoolHandle definitions: This pull request resolves several definition issues in the tf.raw_ops.ExperimentalThreadPoolHandle function, improving its correctness. It excludes changes to other functions related to the name of container and addresses findings from issue #100522.
• pull/100523

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.

1. I'm a spammer and I spam
   • Toxicity Score: 0.75 (Rapid escalation, suspicious behavior, defensive responses)
   • This GitHub conversation involves a user with a suspicious username submitting a commit with unclear intentions, which triggers concern and skepticism among other participants. The tone is cautious and somewhat tense as contributors question the legitimacy and purpose of the submission, leading to a guarded and defensive exchange. The conversation does not progress towards resolution and remains unresolved, with participants wary of potential spam or malicious activity.

---

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
ILCSFNO	12	7	11	8
mihaimaruseac	0	0	0	19
zvoicu000	5	3	0	3
adi1220	4	1	0	5
No author found	9	0	0	0
Venkat6871	0	0	0	9
khteh	0	0	6	2
Ma-gi-cian	5	1	0	1
DottsGit	5	1	0	0
apach301	2	2	2	0