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 a breaking change in the tf.lite API, where the tf.lite.Interpreter Python API is deprecated and moved to ai_edge_litert.interpreter ahead of its removal in version 2.20, alongside updates to C++ constants for better API compatibility. Additionally, the tfl.Cast operation now supports bfloat16 in the runtime kernel, and the libtensorflow packages are no longer published separately but remain accessible via the PyPI package.

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. [STAT:CONTRIBUTION WELCOME] [TYPE:SUPPORT] [COMP:CORE] [TF 2.18] How do we silence noisy messages?: This issue addresses the problem of noisy error messages generated by TensorFlow's CheckNumerics operation, which flood the standard error output with redundant logs that hinder effective debugging. The user is seeking a way to silence or reduce these messages, ideally through a filter, environment variable, or code modification, to improve the clarity of error reporting without losing critical information.

   • The comments discuss various approaches to silencing these messages, including redirecting stderr to null, setting an environment variable to suppress C++ logs, and using an alternative TensorFlow operation that avoids the noisy logs; community members also propose a code change to demote the log level of the redundant error messages to reduce noise while preserving debug information.
   • Number of comments this week: 2

2. [TYPE:DOCS-BUG] [TYPE:BUG] [TF 2.19] tf.keras.Model have not attribute : submodules but the document still use the error attribute: This issue reports a discrepancy between the TensorFlow documentation and the actual API behavior, where the example in the guide uses an attribute submodules on a tf.keras.Model that does not exist, leading to an AttributeError. The user provides a minimal reproducible example showing the error and highlights that the guide incorrectly uses tf.Module instead of subclassing from tf.keras.Model, which causes the missing attribute issue.

   • The comments clarify that the guide example should subclass from tf.keras.Model to access the submodules attribute, provide a corrected code snippet that works without error, and discuss the submission of a pull request to update the documentation accordingly.
   • Number of comments this week: 2

3. [STAT:AWAITING RESPONSE] [TYPE:BUG] [COMP:KERAS] [COMP:OPS] [TF 2.19] Invalid class_weight silently ignored in model.fit(), no error raised: This issue reports that when an invalid class_weight dictionary with non-integer keys is passed to model.fit() in TensorFlow 2.19.0, no error or warning is raised, causing the training to proceed silently without applying the intended class weights. This behavior deviates from expected functionality where such invalid inputs should trigger a ValueError or KeyError, potentially leading to unnoticed misconfigurations during model training.

   • The comments indicate that a pull request addressing the issue was submitted and reviewed, with confirmation that the problem is resolved in the latest tf-nightly builds; contributors expressed interest in fixing the issue, and follow-up improvements related to class weight validation and preprocessing layers were also discussed.
   • Number of comments this week: 2

4. [TYPE:BUG] [COMP:OPS] [AWAITING PR MERGE] [TF 2.19] Fix the axis issue in func tf.raw_ops.Dequantize(): This issue addresses a bug in the tf.raw_ops.Dequantize() function related to the handling and documentation of the axis parameter, where certain axis values produce unexpected errors or behavior despite the input tensor's dimensions. The reporter highlights inconsistencies in axis validation, suggests clearer documentation on the purpose and constraints of the axis argument, and provides reproducible code demonstrating the problem.

   • The comments confirm the issue is reproducible on multiple TensorFlow versions, share a reference gist, announce a related pull request has been opened, and express appreciation for the contribution while encouraging further community involvement and collaboration to finalize the fix.
   • Number of comments this week: 2

5. [STAT:CONTRIBUTION WELCOME] [TYPE:PERFORMANCE] [COMP:CORE] [TF 2.9] Slow iteration/converting of tensors to vector via Python C API: This issue reports a significant performance problem when converting TensorFlow tensors to C++ vector<float> using the Python C API, where the operation takes tens of seconds compared to an almost instantaneous conversion when using the numpy() method first. The user provides detailed system information, code to reproduce the issue, and benchmarks across multiple TensorFlow versions, highlighting that the problem persists even in the latest releases and suggesting that the conversion method via the Python C API is inefficient.

   • Commenters confirmed the issue across various TensorFlow versions and environments, shared reproducible code snippets, and noted the problem remains unresolved after several years; maintainers acknowledged the issue but marked it for community contributions without prioritizing a fix, encouraging users to submit pull requests if interested.
   • Number of comments this week: 1

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: 28

Summarized Issues:

• TensorFlow Lite Memory Safety Vulnerabilities: TensorFlow Lite has multiple memory safety issues including a stack out-of-bounds write in BroadcastBinaryOpSimple when handling tensors with rank > 8, and a heap-buffer-overflow triggered by malformed .tflite models during tensor allocation in the reshape operator. These vulnerabilities can lead to memory corruption or denial of service before inference is run.
• issues/124845, issues/124982

• TensorFlow Lite Android Insufficient Information: A user reported a problem related to TensorFlow Lite on Android but did not provide sufficient system information, reproducible code, or logs needed to diagnose and address the issue. This lack of detail prevents effective troubleshooting.
• issues/124683

• Numerical Overflow and Incorrect Results in Math Functions: Several TensorFlow math functions incorrectly overflow or return invalid results for large float64 inputs, including tf.math.bessel_i0 overflowing to infinity at 713.0, tf.math.erf returning NaN instead of saturating to ±1, and tensorflow.experimental.numpy.hypot overflowing to infinity instead of finite hypotenuse values. These issues cause discrepancies with expected mathematical behavior and other libraries.
• issues/124772, issues/124773, issues/124774

• Incorrect Higher-Order Derivatives in Activation Functions: The second derivatives of tf.nn.elu, tf.nn.selu, and tf.nn.silu functions return incorrect zero values near zero or for negative float64 inputs, indicating bugs in TensorFlow's automatic differentiation higher-order backward pass. These errors affect the accuracy of gradient-based computations involving these activations.
• issues/124830, issues/124831, issues/124832

• Incorrect First-Order Gradients in Activation Functions: The first derivatives (gradients) of tf.nn.elu, tf.nn.selu, and tf.nn.silu return zero or incorrect values for large negative float64 inputs, where finite nonzero gradients are expected. This indicates problems in the backward pass implementations of these activation functions, potentially impacting training stability.
• issues/124833, issues/124834, issues/124835

• Incorrect Gradient for tf.math.sqrt at Minimum Positive Float64: The gradient of tf.math.sqrt returns infinite values for the smallest positive float64 input, despite the analytic derivative being finite, indicating a backward pass bug in the autodiff implementation.
• issues/124836

• JIT Compilation and Autoclustering Inconsistencies and Silent Failures: TensorFlow exhibits inconsistent behavior between eager execution, explicit jit_compile=True, and autoclustering modes, including silent incorrect results or ignored errors for invalid inputs such as output dtype mismatches in tf.map_fn, shape inference errors in tf.reshape with multiple -1 dimensions, nested tuple handling in tf.tuple(), and invalid device placement on nonexistent GPUs. These inconsistencies can cause silent failures or unexpected behavior in compiled graphs.
• issues/124872, issues/124877, issues/124879, issues/124880

• GPU and XLA Runtime Crashes and Illegal Memory Access: TensorFlow 2.20/2.21.0 has critical bugs causing crashes such as CUDA illegal memory access errors during GPU broadcasting operations with very large tensors, segmentation faults in XLA CPU runtime when using negative indices in tf.strided_slice, and fatal assertion failures during oneDNN/MKL layout graph optimization passes triggered by invalid strides or attribute lengths. These bugs cause process aborts and segmentation faults instead of catchable errors.
• issues/124949, issues/124950, issues/124952, issues/124953

• Fatal Crashes from Invalid Configuration or Input Values: TensorFlow crashes fatally when given invalid configuration or input values, including negative inter_op_parallelism_threads causing assertion failure in tf.distribute.Server, NaN or infinite values in boxes tensor causing segmentation fault in tf.raw_ops.CropAndResizeGradBoxes, and slicing bounds errors in tf.raw_ops.BlockLSTMGrad causing aborts. These issues highlight missing input validation and error handling.
• issues/124954, issues/124955, issues/124974

• Shape and Gradient Computation Bugs in JIT-Compiled Graphs: The UnsortedSegmentSum operation incorrectly resolves bounded-dynamic dimensions to their upper bounds during gradient computation under jit_compile=True, causing static shape check failures and invalid argument errors in XLA-compiled graphs. This bug affects shape inference and gradient correctness in compiled functions.
• issues/124963

• Memory Growth and Segmentation Faults from Input Handling: Functions decorated with @tf.function retrace and cache graphs for each new input shape without garbage collection, causing unbounded memory growth and eventual out-of-memory crashes. Additionally, calling .numpy() or string formatting on outputs of tf.repeat with deeply nested single-element tuples causes segmentation faults during EagerTensor to NumPy conversion, indicating insufficient input validation and error handling.
• issues/124972, issues/124973

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: 25

Summarized Issues:

• Memory leaks and resource management issues: Multiple issues report memory leaks or improper resource handling in TensorFlow components. These include leaks when creating tf.constant objects repeatedly in a loop, TensorFlow Lite FFT kernel buffer leaks during failed evaluations, and failure to create empty files with tf.io.gfile.GFile unlike Python's open function.
  • issues/68196, issues/123387, issues/45171

• XLA JIT compilation bugs and inconsistencies: Several issues highlight bugs and unexpected behaviors when using XLA JIT compilation. Problems include crashes with empty inputs in tf.image.non_max_suppression, incorrect int8 tensor operation results due to overflow mishandling, failure to broadcast dynamic shapes, and failure to normalize negative axis indices in tf.transpose under jit_compile.
  • issues/117245, issues/118178, issues/122055, issues/124303

• Crashes and aborts due to invalid inputs or environment issues: Multiple bugs cause TensorFlow to crash or abort instead of raising proper exceptions. These include segmentation faults with large values in tf.raw_ops.CropAndResize, abort traps on MacOS ARM when importing TensorFlow and PyArrow, aborts on zero-width or zero-height inputs to tf.io.encode_png, and crashes when SparseTensor dense_shape contains negative dimensions.
  • issues/100375, issues/98563, issues/124951, issues/107976

• Documentation and feature request for clearer usage examples: There is a request for improved documentation and examples regarding the use of RunOptions such as report_tensor_allocations_upon_oom, as current references are unclear and raise questions about components like config_pb2.
  • issues/17076

• Quantization and data type mismatches: Issues report problems with quantization where input and weight data types mismatch, such as in transpose_conv operations where weights are quantized to int8 but inputs remain float32, causing invalid results.
  • issues/105281

• Build and configuration failures: There are build failures due to CUDA configuration loading errors and incorrect FlatBuffers version override documentation, with the latter also failing due to hardcoded version checks in the build system.
  • issues/123789, issues/124399

• TensorFlow operation argument and behavior inconsistencies: Some operations have inconsistent behavior or error messages, such as tf.raw_ops.MaxPoolGradWithArgmax only supporting int64 despite documentation stating int32 is allowed, and tf.math.reduce_min lacking deterministic tie-breaking options in gradient computation.
  • issues/82837, issues/97688

• Bugs in model training and multi-GPU setups: Training failures occur with NaN losses when using certain pairs of NVIDIA GTX 1080 Ti GPUs in multi-GPU setups, causing training to fail unexpectedly.
  • issues/105552

• Bugs related to symbolic tensors and initializers: XLA JIT compilation fails when Keras initializers use dynamic shapes with symbolic tensors because initializers require concrete integer values, leading to TypeErrors during compilation.
  • issues/105334

• TensorFlow API and layer inconsistencies: The tf.keras.layers.Conv1DTranspose layer works with symbolic Keras inputs but fails with TensorFlow tensors using the same parameters, possibly due to ONEDNN convolution forward propagation errors.
  • issues/113330

• Profiling and timing measurement discrepancies: There is a discrepancy where a custom timer inside a TensorFlow operator reports significantly lower execution time than the full TensorFlow trace during profiling, indicating inconsistent timing measurements.
  • issues/104748

• Control dependency graph issues: The CreateControlDependencies function emits redundant and address-dependent control edges due to an incomplete reachability map, causing over-serialization of concurrent CollectiveReduce operations on a device.
  • issues/124405

• Integer overflow and undefined behavior in TFLite backend: Performing element-wise division of INT32_MIN by -1 in the TensorFlow Lite reference backend causes signed integer overflow and crashes due to undefined behavior.
  • issues/124409

• Tensor slicing and conversion bugs in TFLiteConverter: Slicing a tensor with a Python slice using a literal end index near INT64_MAX results in truncated output during model conversion, whereas an open-ended slice produces correct output.
  • issues/123997

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: 28

Key Open Pull Requests

1. Fix stack overflow in DebugFileIO::RecursiveCreateDir for relative debug dump paths (#123114): This pull request fixes a stack overflow caused by infinite recursion in DebugFileIO::RecursiveCreateDir when handling relative, slash-less debug dump paths by adding a base case that stops recursion on empty directory strings, thereby preventing segmentation faults when loading SavedModels with certain debug operations using relative file URLs.

• URL: pull/124609

• Associated Commits: 1a88e, 0170c, 8656b

2. Use compiler-rt instead of libgcc on s390x: This pull request addresses missing runtime symbols in libgcc on the s390x architecture by configuring Bazel to link against compiler-rt instead of libgcc, ensuring that the necessary symbols (__extendhfsf2 and __truncsfhf2) are available and preventing test failures.

• URL: pull/124782

• Associated Commits: e73bc, 65604

3. Fix ELU gradient precision for negative inputs: This pull request fixes the precision of the ELU activation function's gradient for negative inputs by computing the derivative directly from the input using exp(min(x, 0.0)) instead of reconstructing it from the output, updates GradientTape metadata to retain the ELU input during eager execution, and adds regression tests for both float32 and float64 in graph and eager modes to ensure accurate derivative calculations.

• URL: pull/124902

• Associated Commits: 39ab7, d9d72

Other Open Pull Requests

• Bug fixes for domain and input validation in math and tensor operations: Multiple pull requests address bugs related to input validation and domain checks in TensorFlow operations. These include fixing tf.math.igamma to return NaN for invalid inputs, rejecting scalar inputs in XLA TensorArrayScatterOp, validating axis values in Quantize/Dequantize kernels, rejecting INT64_MIN axis in GatherV2, and adding input size validation in tf.raw_ops.Where to prevent crashes from large tensors.
• [pull/124927, pull/124930, pull/124760, pull/124935, pull/124931]

• Improvements and fixes in numpy compatibility and experimental numpy functions: Several pull requests fix issues in TensorFlow's experimental numpy module to align behavior with numpy standards. These include fixing linspace and logspace to honor explicit dtypes, correcting unary math functions to avoid AttributeErrors without numpy behavior enabled, and fixing tnp.isinf and related functions to handle unconverted inputs and non-floating dtypes properly.
• [pull/124932, pull/124957, pull/124958]

• Numerical stability and overflow fixes in special math functions: Pull requests improve numerical stability and prevent overflow in special math functions. Fixes include a numerically stable implementation of tensorflow.experimental.numpy.hypot for large float64 inputs, saturation of tf.math.erf outputs for large inputs to prevent NaNs, and a stable log-space computation for tf.math.bessel_i0 to avoid exponential overflow.
• [pull/124888, pull/124889, pull/124890]

• Memory safety and out-of-bounds prevention in matrix and tensor operations: Fixes focus on preventing out-of-bounds memory accesses and segmentation faults. These include validating permutation indices in SparseMatrixSparseCholesky, fixing negative shape dimension checks in RaggedTensorToTensorOp, and reconciling bounded-dynamic dimensions in BCastGradArgsOp to prevent erroneous broadcast failures.
• [pull/124996, pull/124995, pull/124966]

• Logging and error message improvements: One pull request reduces noisy error output by demoting GPU CheckNumerics kernel anomaly detection logs from error to verbose level, aligning with CPU and V2 GPU implementations while preserving debugging information.
• [pull/125002]

• Regression tests and validation enhancements: Several pull requests add regression tests and improve validation to prevent future bugs. These include tests for XLA fake-quant rounding behavior, validation of ksize and strides lengths in quantized pooling ops, and tests ensuring consistent rejection of nested structures in tf.tuple.
• [pull/124718, pull/124959, pull/124931]

• Fixes for TensorList and tensor broadcasting issues: Fixes include validation of element dtype mismatches in XLA TensorList read kernels to prevent garbage data returns, and replacing fixed-size broadcast metadata arrays with dynamically growing vectors in TFLite to prevent stack overflow on high-rank broadcasts.
• [pull/124874, pull/124904]

• Documentation and formatting improvements: Minor improvements to documentation files include adding a header and enhancing formatting in ISSUES.md and fixing punctuation errors in CONTRIBUTING.md to improve readability.
• [pull/124870, pull/124871]

• Miscellaneous updates: A pull request that updates some files in the TensorFlow project without further specification.
• [pull/124926]

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: 52

Key Closed Pull Requests

1. Fix: Prevent kernel crash on duplicate TypeSpec registration in notebooks: This pull request addresses the issue of TensorFlow kernel crashes caused by re-running cells that register the same TypeSpec class in interactive notebook environments by updating the type_spec_registry.py to detect and safely handle duplicate registrations of the exact same class, logging a warning instead of raising a hard ValueError, thereby improving developer experience without compromising the safety of the registry.

• URL: pull/121669

• Associated Commits: bed2f, adffd, 540b5, 76b5c, 9d6fa, 18ac1, 6df1a, c3a94, 1a72d, c6688, 94ffb, 7c91a, 4a221, 0d928, 9b7ce, 2b693, 66a78, a2b5b

2. Fix the dealt of minlength and maxlength in tf.math.bincount: This pull request addresses and attempts to fix issues related to the handling of the minlength and maxlength parameters in the tf.math.bincount function in TensorFlow, as indicated by the title and associated commits, but it was not merged.

• URL: pull/99514

• Associated Commits: 931f9, 96cb6, 71626, b68de, 6fc15, 7f4a4

3. Fix segfault in AudioSpectrogram with negative window/stride. Fixes #108664: This pull request fixes a segmentation fault in the AudioSpectrogram operation caused by negative window size or stride values by adding validation checks in the code and includes new test cases to verify these fixes.

• URL: pull/108808

• Associated Commits: a09ba, 48c7b, 65c13, 00812, 0d724

Other Closed Pull Requests

• NumPy API updates: These pull requests replace deprecated and removed NumPy aliases and APIs such as np.float, np.object, np.str, np.cast, and np.string_ with supported equivalents to maintain compatibility with NumPy versions 1.24 and 2.x. This ensures the TensorFlow codebase remains up-to-date with current NumPy standards.
  [pull/117688]

• Crash and segmentation fault fixes: Multiple pull requests fix crashes caused by issues such as partially-decoded Variant objects, invalid resource handles, and empty input tensors in XLA lowering of NonMaxSuppression. These fixes include adding validation checks and proper clearing of objects to prevent segmentation faults and ensure clean error handling.
  [pull/123829, pull/113269, pull/124047]

• Integer overflow and buffer overflow prevention: Several pull requests address integer overflow vulnerabilities and potential heap buffer overflows by adding integer overflow checks, non-negativity validations, and overflow-safe arithmetic operations in TensorFlow Lite detection postprocessing and buffer bounds checks. These changes prevent unsafe buffer allocations and out-of-bounds writes.
  [pull/112654, pull/122386, pull/123910]

• TensorArray dynamic growth and XLA compatibility: A pull request implements dynamic growth support for TensorArray with an initial size of zero under XLA while loops by fixing TensorArraySizeV3 to correctly return the live leading dimension. This ensures stack operations match eager execution and includes regression tests for the fix.
  [pull/119383]

• Code modernization and formatting improvements: A pull request updates legacy percent-formatting print statements to modern Python f-strings with the !r conversion flag in a shell script, improving consistency and modernizing string formatting without changing functionality.
  [pull/121676]

• Validation and error handling improvements in operations: Pull requests add validation checks to operations such as MatrixDiagOp and Conv-Transpose layers to prevent hard failures and improve error messages. These include ensuring diagonal tensors have appropriate dimensions and validating unsupported stride and dilation_rate combinations.
  [pull/110850, pull/121735]

• Windows GPU warning fix: A pull request fixes an issue where the Windows GPU deprecation warning was incorrectly triggered on internal Keras calls by adding a check to show the warning only when the user explicitly queries GPU devices.
  [pull/112008]

• Documentation fixes and clarifications: Multiple pull requests improve documentation by adding missing imports, fixing broken links, clarifying deterministic behavior in reduce_min/max gradients, and correcting argmax type support in MaxPoolGradWithArgmax. These changes enhance usability and prevent user confusion without affecting functionality.
  [pull/124018, pull/108761, pull/109337, pull/111931]

• Numerical stability fixes: A pull request fixes the gradient of tf.norm to prevent NaN or Inf values when the input is zero or very small by modifying the Euclidean norm gradient calculation. Another addresses floating-point rounding issues in the erfinv function near the float32 upper boundary.
  [pull/108761, pull/121735]

• XLA compilation and lowering fixes: Pull requests fix XLA compilation failures in tf.experimental.numpy.compress by adjusting tensor slicing and prevent segmentation faults in XLA lowering of NonMaxSuppression by handling empty input tensors properly. These ensure successful compilation and runtime stability under XLA.
  [pull/122563, pull/124047]

• Regression test additions: Several pull requests include regression tests to verify fixes for issues such as TensorArray dynamic growth, detection postprocess overflow checks, and NaN propagation in XLA kernels, ensuring robustness and preventing regressions.
  [pull/119383, pull/122386, pull/117046]

• Code cleanup and minor corrections: Pull requests remove obsolete macros and static genrule wrappers related to CUDA data directories and fix minor punctuation issues in documentation files to improve readability and maintainability.
  [pull/124796, pull/124837]

• Security warnings and documentation: A pull request adds security warnings to the tf.io.gfile documentation to inform users about the lack of path validation and sandboxing, emphasizing the need for applications to validate file paths from untrusted sources.
  [pull/111090]

• Tutorial and example fixes: A pull request fixes runtime issues in the mnist_tflite.py tutorial by updating deprecated APIs, correcting accuracy computation, and adding regression tests for prediction logic.
  [pull/119440]

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.