Weekly GitHub Report for Tensorflow - 2024-07-15 12:00:54
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.
I. Issues
1.1 Open Issues
Open Issues This Week: 18
Summarized Issues:
- TensorFlow Lite Model Issues: Users have encountered various issues when working with TensorFlow Lite models, including setting input tensors after post-training quantization, missing C header files in the AAR version, and unsupported operations in Flutter applications. These issues often result in compilation errors, null values during inference, and difficulties integrating with the latest TensorFlow versions. Users have tried multiple solutions, such as creating .aar files and using delegates, but continue to face challenges.
- Compilation and Build Errors: Several users have reported encountering compilation and build errors while working with TensorFlow, including issues with custom operators, TensorFlow 2.18.0 on CentOS 9 Stream, and TensorFlow Lite for ARM64 on Ubuntu 22.04. These errors often involve specific files and functions, such as
Status::OK(), lambda function conversion, and multiple redefined warnings. Users are seeking guidance on resolving these errors to successfully compile and build their projects.
- Model Loading and Conversion Errors: Users have faced issues when loading and converting TensorFlow models, including
ValueErrordue to missingsaved_model.pbfiles,AttributeErrorwhen converting Keras models to TensorFlow Lite, and errors in TFLite-Model-Maker due to deprecated NumPy attributes. These errors disrupt the workflow and prevent successful model deployment and usage.
- Performance and Compatibility Issues: Users have reported performance and compatibility issues, such as the feasibility of using custom GPU external delegates with TensorFlow Lite, discrepancies in Conv1D layer results on macOS versus Linux, and LSTM training getting stuck on Mac OS. These issues affect the efficiency and reliability of TensorFlow models across different platforms and configurations.
- Functionality Requests and Bugs: Users have requested new functionalities, such as the addition of a sparsemax function for TabNet, and reported bugs, such as negative results from the binary cross-entropy function and memory issues when installing tflite-model-maker. These requests and bugs highlight the need for continuous improvements and updates to the TensorFlow framework.
- Custom Training and Model Issues: Users have encountered issues with custom training loops and specific models, such as
UnliftableErrorwithtf.distribute.MirroredStrategyand problems creating a BERTQA model with tflite-model-maker. These issues often involve tensor dependencies and evaluation failures, hindering the development and deployment of custom models.
1.2 Top 5 Active Issues:
We consider active issues to be issues that have generated much discussion in the issue's comments.
- TF 2.17.0 RC0 Fails to work with GPUs (and TF 2.16 too): This issue is about TensorFlow version 2.16.1 failing to detect and utilize GPUs on Linux systems, even when installed with the
tensorflow[and-cuda]package. Users report that despite following the official installation instructions, TensorFlow does not recognize the GPU, and additional manual steps are required to set environment variables and create symbolic links to the necessary CUDA libraries.
- The comments section reveals that many users are experiencing similar issues, with some finding workarounds by manually setting environment variables or downgrading to TensorFlow 2.15.1. There are also discussions about the upcoming TensorFlow 2.17.0 release, with some users testing the release candidate and reporting mixed results. The TensorFlow team is urged to address these issues in future releases and update the official documentation to include the necessary steps for proper GPU utilization.
- Number of comments: 158
- Bug - pip install tflite-model-maker fails or hangs forever: This issue is about the failure or indefinite hanging of the
pip install tflite-model-makercommand in Google Colab, which prevents users from installing the package necessary for TensorFlow Lite model making. The problem persists across different attempts and configurations, including using nightly builds and various dependency management strategies, leading to errors related to package dependencies and version conflicts.
- The comments section reveals that many users are experiencing the same issue, with some reporting that the installation process fills up the storage space and others encountering dependency conflicts. Various workarounds, such as downgrading Python versions or using alternative installation methods, have been suggested but with limited success. The TensorFlow team has acknowledged the issue and recommended using the MediaPipe Model Maker as an alternative solution.
- Number of comments: 135
- cuDNN, cuFFT, and cuBLAS Errors: This issue is about errors and warnings encountered when running a GPU test with TensorFlow, specifically related to the inability to load cuDNN, cuFFT, and cuBLAS libraries, which affects the speed of training neural networks. The user is running TensorFlow on WSL2 Linux Ubuntu 22 with an NVIDIA Geforce GTX 1660 Ti GPU and has tried various solutions without success.
- The comments discuss various attempts to resolve the issue, including reinstalling TensorFlow with the
tensorflow[and-cuda]package, downgrading TensorFlow versions, and adjusting CUDA and cuDNN installations. Some users report partial success or workarounds, while others continue to face the same errors. The issue appears to be related to TensorFlow's handling of CUDA libraries and NUMA support in WSL2.- Number of comments: 126
- How can I clear GPU memory in tensorflow 2?: This issue is about a user experiencing difficulties with TensorFlow 2.1.0 where GPU memory is not being released after model training, causing subsequent models to fail due to out-of-memory errors. The user has tried various solutions found online, but none have worked, and they are seeking a way to clear the GPU memory without restarting their computer.
- The comments section includes multiple suggestions such as using
tf.keras.backend.clear_session(), limiting GPU memory growth, running training in separate processes, and using thecuda_malloc_asyncallocator. Despite these suggestions, many users report that the issue persists, and some have resorted to using subprocesses or switching to other frameworks like PyTorch.- Number of comments: 120
-
Feature Request: GPUOptions for Go binding: This issue is a feature request to add GPUOptions for the Go binding in TensorFlow, as the current implementation does not allow specifying options for
NewSession. The problem is that the GPUOptions struct is in an internal package, andgo generatedoes not work for the protobuf directory, making it impossible to specify GPUOptions.- The comments discuss the need for a Go API to configure TensorFlow sessions, with some users proposing solutions and creating pull requests. Despite initial progress, the proposed changes were reverted, leading to confusion and multiple users asking if the issue is still open and how they can contribute. The thread also includes many off-topic comments and requests for guidance on contributing to open source projects.
- Number of comments: 95
1.3 Top 5 Quiet Issues:
We consider quiet issues to be issues that have been opened in this project for the longest time. The team should work together to get these issues resolved and closed as soon as possible.
-
Make TensorFlow compatible with PyPy: This issue is about making TensorFlow compatible with PyPy, which is a Python interpreter known for its speed and efficiency. The author acknowledges that this is not a priority and anticipates it will take a significant amount of time to achieve, but believes it would be a valuable enhancement.
- Open for 3162 days, 17 hours, 51 minutes
-
tf.dynamic_stitchgradient is incorrect: This issue highlights a problem with the gradient calculation of thetf.dynamic_stitchfunction in TensorFlow, where the analytic gradient does not match the expected numeric gradient. Specifically, the issue arises when there are duplicate indices being merged, leading to an incorrect analytic gradient that does not reflect the fact that the inputxis completely overwritten by a constant.- Open for 2712 days, 06 hours, 48 minutes
-
Numerical instability of gradient calculation of tf.norm (nan at 0, inf for small values) : This issue highlights a numerical instability in the gradient calculation of the
tf.normfunction in TensorFlow, where the gradient results innanat zero values andinffor extremely small values. The problem is demonstrated with a minimal example and was encountered in a real-world scenario while implementing a custom loss function, causing incorrect gradient values when two embeddings were very close to each other.- Open for 2533 days, 18 hours, 26 minutes
-
Support Windows builds through clang: This issue is about adding support for building TensorFlow on Windows using the clang compiler, as currently, the tested configurations for Windows use MSVC. Additionally, it discusses the potential challenges and benefits of enabling cross-building for Windows from a Linux environment, particularly concerning the CUDA SDK installation.
- Open for 2439 days, 02 hours, 18 minutes
-
Feature request: Allow the build to use the system-installed protobuf lib: This issue is a feature request to allow the TensorFlow build process to utilize the system-installed protobuf library instead of the default
protobuf_archiveversion. The request stems from a problem encountered while packaging TensorFlow for Arch Linux, where symbol conflicts arise when users attempt to use both protobuf and TensorFlow in a binary due to conflicting protobuf symbols.- Open for 2374 days, 07 hours, 35 minutes
1.4 Closed Issues
Closed Issues This Week: 23
Average Issue Close Time (This Week): 433.32 days
Summarized Issues:
- Migration to TensorFlow 2.0 API: Migrating libraries and models to TensorFlow 2.0 involves converting existing code, running unit tests, and ensuring compatibility with new features like eager execution and distribution strategies. This process includes updating the Tensor2Tensor library and sequence-to-sequence models for Neural Machine Translation (NMT). Ensuring all necessary tests and engineering artifacts are included is crucial for a smooth transition.
- Memory Leak Issues: A memory leak was identified in TensorFlow 1.12 when using
report_tensor_allocations_upon_oomwithembedding_lookup, which does not occur in TensorFlow 1.10 or when usingtf.gather. This issue highlights the importance of identifying and resolving memory management problems in different TensorFlow versions. Proper handling of tensor allocations is critical to avoid out-of-memory errors.
- Documentation and Source Code Links: There is a need for TensorFlow's Python API documentation to include links to the corresponding C++ source code. The current documentation only points to the Python code, making it difficult for users to trace operations that are thin wrappers around C++ implementations. Additionally, broken links in the Keras documentation lead to 404 errors, complicating the user experience.
- Security Fix Identification: Users have difficulty locating specific commits related to the security fix for CVE-2022-23569 in TensorFlow. The advisory mentioned multiple commits, but they are hard to identify due to the common occurrence of "Check-fails" in the codebase. Clear documentation and commit tracking are essential for security transparency.
- TPU and GPU Compatibility Issues: Using the
groupsparameter in thekeras.layers.Conv2DAPI on a TPU-VM device results in aNotFoundErrordue to the absence of a registered transfer manager for the platform. This issue does not occur on a GPU, highlighting the need for consistent API behavior across different hardware platforms.
- Segmentation Faults and Model Conversion Issues: Segmentation faults occur when reading tensor details with sparsity using the TFLite Python interpreter on specific TensorFlow and Python versions. Additionally, models created with TensorFlow 2.15.0 or earlier can be converted to TFLite, but models created with TensorFlow 2.16.0rc0 fail to convert, indicating incompatibilities between Keras versions 2 and 3.
- Quantization and Inference Issues: TensorFlow's "per-channel" quantization incorrectly sets the quantized dimension to the batch axis instead of the channel axis, resulting in incorrect inference outputs. Accurate quantization is crucial for model performance and correctness, especially in resource-constrained environments.
- Null Pointer Dereference: A potential null pointer dereference in the
mlir::odml::MatchInverseScalesOperand()function was identified by a static analyzer. The issue involves an undefined variable being printed in the error message, which needs to be addressed to prevent runtime errors.
- Memory Errors on Mobile Devices: An Android Tflite model fails to load on a Samsung S23 device using a GPU Delegate due to an "Out of host memory" error. This issue is likely caused by exceeding the Dalvik-heap memory limit, requiring troubleshooting and alternative solutions for memory management on mobile devices.
- Installation and Compatibility Issues: Installing the
tflite-model-makerpackage in a Python virtual environment fails due to an AttributeError related to thepkgutilmodule. Additionally, GCS operations usingtf.io.gfilefail in TensorFlow versions 2.17.0rc0 and 2.18 nightly when not running on a GCP virtual machine, causing authentication and HTTP request errors.
- Deprecation Warnings and Compilation Errors: A deprecation warning occurs when importing TensorFlow due to the deprecation of the
xla_computationAPI in JAX. Additionally, a compilation error in TensorFlow version 2.16.1 on a Linux Ubuntu 22.04 system is caused by unhandled enumeration values in the DNNL library.
- Model Loading and Conversion Errors: TensorFlow is unable to load a previously saved model and convert it to ONNX, resulting in an "AttributeError: '_UserObject' object has no attribute 'add_slot'" error. This issue highlights the need for robust model serialization and deserialization processes.
- Performance Optimization: Users experience significantly increased training times for their TensorFlow models after upgrading from TF 1.x to Keras 2 and later to TF 2.13, despite upgrading their hardware. Troubleshooting efforts focus on optimizing performance and CPU utilization to achieve efficient model training.
- Custom Delegate Integration: Guidance is sought on integrating and using a custom TensorFlow Lite stable delegate, compiled as a shared library (.so file), within a C++ application. Ensuring the delegate is properly loaded and the model is executed by it is crucial for leveraging custom optimizations.
- Platform-Specific Code Issues: Inaccurate shoulder width and height measurements on iOS devices using a React Native and Expo application are reported, where the code works correctly on Android but fails on iOS. Handling platform-specific code for TensorFlow on iOS is necessary to ensure consistent results across devices.
- TypeError in Custom Metrics: A
TypeErroroccurs when usingtf.data.Dataset.from_generatorin TensorFlow, where aNoneTypeobject is passed to a custom metric function. This error is due to unsupported comparison betweenNoneTypeandint, requiring proper handling of data types in custom metrics.
- Spam and Unauthorized Content: The TensorFlow GitHub project faces issues with spam, including posts promoting services for hacking Instagram accounts. Efforts to remove spam and ensure the integrity of the project are ongoing, with requests for reproducible code to better understand and address the problem.
1.5 Issue Discussion Insights
This section will analyze the tone and sentiment of discussions within this project's open issues within the past week to identify potentially heated exchanges and to maintain a constructive project environment.
-
Bug - pip install tflite-model-maker fails or hangs forever
- Toxicity Score: 0.55 (Frustration, repeated issue confirmations, urgency due to deadlines)
- This GitHub conversation involves multiple users discussing an issue with installing the
tflite-model-makerpackage. The conversation starts with user Huxwell reporting the problem and providing detailed error logs. Other users, such as UsamaHameed1 and synandi, join in to confirm they are experiencing the same issue. The tone is initially collaborative, with users sharing their attempts to resolve the problem. However, frustration builds as the issue persists, and some users express urgency due to project deadlines. Tension arises when users like juanjricci and mehlkelm emphasize the critical impact on their work. Despite some suggestions and temporary workarounds, the problem remains unresolved, leading to continued expressions of dissatisfaction and requests for a permanent fix.
-
TF 2.17.0 RC0 Fails to work with GPUs (and TF 2.16 too)
- Toxicity Score: 0.65 (Frustration, unresolved issues, technical complexity, calls for better testing and documentation)
- This GitHub conversation involves multiple users discussing issues with TensorFlow 2.16.1 and its compatibility with GPUs. The conversation starts with JuanVargas reporting a bug where TensorFlow fails to recognize GPUs. Various users, including DiegoMont, Shayan Shahrokhi, and others, share their experiences and potential solutions, such as adjusting environment variables and downgrading TensorFlow versions. The tone is generally collaborative, but there are moments of frustration, particularly when suggested fixes do not work. The conversation also includes technical discussions about CUDA and cuDNN versions, and the need for updated documentation. Despite some successful workarounds, the issue remains unresolved for many, leading to calls for better testing and documentation from the TensorFlow team.
-
- Toxicity Score: 0.55 (Frustration, persistent unresolved issue, informal language)
- This GitHub conversation involves multiple users, including Rajcr2, DineshNeupane, and devVegaAn, discussing a persistent issue with TensorFlow and Keras compatibility. The conversation starts with Rajcr2 reporting an error, followed by various users suggesting potential solutions and workarounds. The tone is generally collaborative, with users offering detailed advice and asking for further clarification. However, there are moments of frustration, particularly when suggested solutions do not resolve the issue, as seen in comments like "shits not working bro." The conversation includes polite requests for help and expressions of gratitude, but also signs of impatience and dissatisfaction as the problem persists.
II. Pull Requests
2.1 Open Pull Requests
Open Pull Requests This Week: 5
Pull Requests:
- TensorFlow Lite quantization consistency: This pull request addresses an issue in TensorFlow Lite where the
broadcast_tooperator's input and output quantization information were not consistent. The inconsistency caused type checking failures during MLIR lowering. The fix enforces the same quantization scale for both input and output using theSameOperandsAndResultsScaleattribute.
- Cluster resolver for OpenPBS, PBSPro, and Torque: This pull request introduces a cluster resolver specifically designed for schedulers based on OpenPBS, PBSPro, and Torque. This enhancement improves the TensorFlow project's compatibility with these systems. It allows for better integration and resource management in environments using these schedulers.
- Auto Mixed Precision (AMP) Optimizer bug fix: This pull request addresses a bug fix to ensure that the Auto Mixed Precision (AMP) Optimizer does not exit when handling float16 operations. The issue was specific to supported Intel E-Core CPU platforms. The fix ensures stability and proper functioning of the AMP Optimizer on these platforms.
- Cherry-picking commits for TensorFlow: This pull request involves cherry-picking the commit "130d583ff62" to remove the
noexceptspecifier from the optimized function graph information. Another pull request aims to cherry-pick the commit a2bda886281, which addresses a fix for a breakage in TensorFlow Serving, into the r2.17 branch. These cherry-picks ensure that critical fixes and optimizations are included in the relevant branches.
2.2 Closed Pull Requests
Closed Pull Requests This Week: 11
Summarized Pull Requests:
- Host Stream Implementation: This pull request adds an empty implementation for the host stream in the XLA StreamExecutor to prevent logging an ERROR message when the SetPriority function is called on the host stream. The implementation ensures that the function call does not result in unnecessary error logs. This change is aimed at improving the logging behavior in the TensorFlow project.
- Code Readability and Performance: This pull request involves rewriting a nested if block into a more readable and potentially more performant nested match block in the TensorFlow project. The change aims to enhance code clarity and maintainability. Additionally, it may offer performance benefits by optimizing the control flow.
- Segmentation Fault Fix: This pull request addresses a segmentation fault issue in the Python
InterpreterWrapperby adding a check fornullptrwhenTensorSparsityParametersencounters Tensors without ablock_map. The fix aims to prevent crashes and improve the stability of the TensorFlow project. This change is specifically targeted at resolving issue #62058.
- Typographical Error Corrections: This pull request addresses and corrects several typographical errors in the TFLite metadata_writer_tutorial.ipynb file within the TensorFlow project. The corrections improve the documentation quality and readability. This change ensures that users have access to accurate and error-free tutorial content.
- Function Optimization: This pull request addresses the optimization of the GetActivationName utility function in the remapper implementation by passing the input string by constant reference. This change avoids the overhead of copying the string, thereby improving performance. The optimization is part of ongoing efforts to enhance the efficiency of the TensorFlow codebase.
- Error Message Improvement: This pull request addresses and resolves issue #67556 by removing a null pointer from an error message in the TensorFlow project. The change aims to improve the clarity and accuracy of error messages. This fix contributes to better debugging and user experience.
- GCS Operations Fix: This pull request addresses the issue of TensorFlow Google Cloud Storage (GCS) operations failing in certain environments by reverting a previous commit. The commit had added a trailing dot to the storage.googleapis.com domain, which caused failures. The revert aims to resolve the problem for the 2.17 release.
- Removal of Distutils Remnants: This pull request addresses the removal of distutils remnants from the TensorFlow project. It references a specific issue and applies a patch from an Arch Linux packaging repository, with credits to the contributor @loqs. This change is part of the effort to modernize the build system.
- Version Update to 2.17.0: This pull request involves updating the version numbers to TensorFlow 2.17.0. It ensures that the relevant files (
core/public/version.h,tools/pip_package/setup.py, andtensorflow/tensorflow.bzl) are correctly modified without any lingering old version strings. This update is crucial for the release process.
- StableHLO Sort Implementation: This pull request introduces the implementation of the stablehlo_sort operation. It includes unit tests for the stablehlo sort, adds a function to convert bfloat16 to TensorFlow Lite type, and incorporates a Comparator Subgraph. These additions enhance the functionality and testing coverage of the TensorFlow project.
- Dummy File Creation: This pull request involves the creation of a dummy file and includes a note to ignore the request. The title "Created Dummy File" and the body "Plz Ignore this request" indicate that it is not intended for actual integration. This request serves as a placeholder or test.
2.3 Pull Request Discussion Insights
This section will analyze the tone and sentiment of discussions within this project's open pull requests within the past week 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 pull requests from the past week.
III. Commits
3.1 Commits
Commits This Week: 303
Summarized Commits:
- Automated Code Changes: Numerous commits involve automated code changes, as indicated by the commit messages and PiperOrigin-RevId references. These changes are likely part of routine maintenance or automated refactoring processes.
- LLVM Integration: Several commits integrate the LLVM project at specific commits, updating the usage to align with changes made in those revisions. This ensures the project remains compatible with the latest LLVM updates.
- Performance Optimizations: Multiple commits focus on optimizing performance, such as reducing CPU operation times, improving the efficiency of the
CopyThunkfor small shapes, and optimizing theRemoveUnusedSymbols/Dimensionsprocess in XLA:GPU.
- Bug Fixes and Issue Resolutions: Various commits address and resolve specific issues, such as fixing race conditions, segmentation faults, build errors, and incorrect sharding propagation. These fixes ensure the stability and correctness of the project.
- Refactoring and Code Cleanup: Many commits involve refactoring code, removing unused functions, parameters, and dependencies, and relocating code to improve maintainability and organization. Examples include refactoring the
SoftmaxRewriterTritonfunction and removing theConvertMlirToGraphdeffunction.
- Test Enhancements and Additions: Several commits introduce new tests or enhance existing ones to ensure correct behavior and improve test coverage. This includes adding tests for the
ifrt client,triton_support_test, andDumpProtobufToFilefunctions.
- XLA:GPU Enhancements: Numerous commits focus on enhancing the XLA:GPU component, such as introducing support for custom call TopK thunk, optimizing the KernelThunk performance, and ensuring correct handling of scatters with non-unique indices.
- ROCm and CUDA Updates: Multiple commits address issues and introduce features specific to ROCm and CUDA, such as adding ROCM-6.2 features, addressing build errors, and updating the CUDA backend to use
libnvjitlink.
- Sharding and Slicing Improvements: Several commits update the sharding propagation logic for dynamic slice and dynamic update slice operations, ensuring correct sharding management and improving performance.
- Backward Compatibility and Forward Compatibility: Various commits update the forward compatibility horizon and GraphDef version to ensure the project remains compatible with future changes and improvements.
- Documentation and Typo Corrections: Some commits focus on updating outdated documentation, correcting typographical errors, and ensuring clearer and more accurate information.
- New Features and Functionalities: Several commits introduce new features, such as the XNNPACK delegate for the
BuiltinOperator_GELUoperation, theThunk::ExecuteSessionfeature, and theEmbeddingLookUpoperation with support for both float and int8 data types.
- Internal Changes and Configurations: Many commits involve internal changes, such as updating Copybara configurations, making internal refactoring changes, and modifying test configurations to improve the project's internal structure and processes.
- Error Handling and Logging Improvements: Various commits enhance error handling and logging, such as introducing type safety to the XLA logging system and providing clearer error messages in the HloVerifier.
- Third-Party Integrations and Updates: Several commits involve integrating third-party libraries and updating dependencies, such as updating the urllib3 library and integrating the Triton library.
- Dynamic Padder and RngBitGenerator: Some commits introduce runtime checks and tests to ensure the
DynamicPadderandRngBitGeneratoroperations are properly expanded, preventing crashes and ensuring correct functionality.
- Memory and Resource Management: Multiple commits focus on improving memory and resource management, such as aligning atomic counters to prevent false sharing and optimizing the
bytes_accessedmetric for assigning IO-bound parallel tasks.
- TensorFlow Serving and TFLite Updates: Several commits address issues and introduce updates specific to TensorFlow Serving and TFLite, such as resolving dependency issues and removing unnecessary dependencies.
- StableHLO and MLIR-HLO Enhancements: Various commits enhance the StableHLO and MLIR-HLO components, such as relocating StableHLO passes and updating the
tf_to_tfl_flatbuffer_helpers.
- Concurrency and Parallelism: Some commits introduce features to manage concurrency and parallelism, such as the
Thunk::ExecuteSessionfeature and optimizing task scheduling overheads for XLA programs with numerous small thunks.
- Backward Compatibility Fixes: Several commits revert previous changes to ensure backward compatibility, such as reverting the removal of support for CUDA versions below 12.3 and restoring the input-output alias configuration.
- Code Simplification and Efficiency: Multiple commits focus on simplifying code and improving efficiency, such as replacing custom functions with LLVM's built-in functions and optimizing the multi-result affine map simplifier.
- New Test Cases and Examples: Some commits introduce new test cases and examples, such as a fundamental test case for a circular pipeline collective permute and a simple example demonstrating how to use the 'shardy' tool.
- Deprecation and Removal of Obsolete Code: Various commits deprecate and remove obsolete code, such as removing the
XLA_FFI_LAZY_DECODED_BUFFERdefine and deprecating parts of thestr_utilmodule that map toabslfunctions.
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, or created at least 1 pull request in the past month.
| Contributor | Commits | Pull Requests | Issues |
|---|---|---|---|
| TensorFlower Gardener | 1116 | 0 | 0 |
| GitHub | 11 | 0 | 0 |
| tensorflow-jenkins | 0 | 7 | 0 |
| Leo-Lifeblood | 0 | 0 | 3 |
| damjandakic93 | 0 | 0 | 3 |
| Mirgahney | 0 | 2 | 0 |
| msteiner-google | 0 | 1 | 1 |
| NobuoTsukamoto | 0 | 1 | 1 |
| MengXiangDing | 0 | 0 | 2 |
| fujunwei | 0 | 0 | 2 |
| ravin00 | 0 | 0 | 2 |
| devapriyas2001 | 0 | 0 | 2 |
| guillaume-tgl | 0 | 0 | 2 |
| hguandl | 0 | 0 | 2 |
| Pol Dellaiera | 1 | 0 | 0 |
| Yimei Sun | 1 | 0 | 0 |
| Michael Hudgins | 1 | 0 | 0 |
| nitins17 | 0 | 1 | 0 |
| quoctruong | 0 | 1 | 0 |
| mattbahr | 0 | 1 | 0 |
| hsharsha | 0 | 1 | 0 |
| MichaelHudgins | 0 | 1 | 0 |
| belitskiy | 0 | 1 | 0 |
| vladbelit | 0 | 1 | 0 |
| psunn | 0 | 1 | 0 |
| yimeisun123 | 0 | 1 | 0 |
| drupol | 0 | 1 | 0 |
| lhutton1 | 0 | 1 | 0 |
| vimalrajv | 0 | 1 | 0 |
| akhilgoe | 0 | 1 | 0 |
| RahulSundarMCW | 0 | 1 | 0 |
| alifbasha22 | 0 | 1 | 0 |
| captainst | 0 | 0 | 1 |
| AnyLifeZLB | 0 | 0 | 1 |
| MAlmuzaini | 0 | 0 | 1 |
| SMH17 | 0 | 0 | 1 |
| Dustglaxy | 0 | 0 | 1 |
| Jerry-yz | 0 | 0 | 1 |
| GreatKampfstern | 0 | 0 | 1 |
| gachombamuthiri | 0 | 0 | 1 |
| johnlarkin1 | 0 | 0 | 1 |
| willendzw | 0 | 0 | 1 |
| ninjaguardian | 0 | 0 | 1 |
| MatthiasNeumueller | 0 | 0 | 1 |
| chiamp | 0 | 0 | 1 |
| myh1234567 | 0 | 0 | 1 |
| hui-li-xf | 0 | 0 | 1 |
| kindaTall | 0 | 0 | 1 |
| BenCrulis | 0 | 0 | 1 |
| H4dr1en | 0 | 0 | 1 |
| maxwellbrien1990 | 0 | 0 | 1 |
| lgeiger | 0 | 0 | 1 |
| Nayana-ibm | 0 | 0 | 1 |
| yanncalec | 0 | 0 | 1 |
| Suryansh-patidar | 0 | 0 | 1 |
| kupix | 0 | 0 | 1 |
| feranick | 0 | 0 | 1 |
| cpappasILMX | 0 | 0 | 1 |
| ThorvaldAagaard | 0 | 0 | 1 |
| cz2h | 0 | 0 | 1 |
| yousifj129 | 0 | 0 | 1 |
| manswad | 0 | 0 | 1 |
| SsomsakTH | 0 | 0 | 1 |
| weebao | 0 | 0 | 1 |
| basi880830 | 0 | 0 | 1 |
| WeichenXu123 | 0 | 0 | 1 |
| NoranNabil | 0 | 0 | 1 |
| mrezaanony | 0 | 0 | 1 |
| kkk-jpeg | 0 | 0 | 1 |
| CV-42 | 0 | 0 | 1 |
| tomneo2004 | 0 | 0 | 1 |
| pekaalto | 0 | 0 | 1 |
| npanpaliya | 0 | 0 | 1 |
| pataneanant | 0 | 0 | 1 |
| suyash-narain | 0 | 0 | 1 |
| hamzaziizzz | 0 | 0 | 1 |
| arpanptnk85 | 0 | 0 | 1 |
| luanft | 0 | 0 | 1 |
| KarenMars | 0 | 0 | 1 |
| Sklaebe | 0 | 0 | 1 |
| Aweptimum | 0 | 0 | 1 |
| alykhantejani | 0 | 0 | 1 |
| baqwas | 0 | 0 | 1 |
| anilkumarvvss | 0 | 0 | 1 |
| asp616848 | 0 | 0 | 1 |
| hiwothadush | 0 | 0 | 1 |
| junwha0511 | 0 | 0 | 1 |
| ramonhollands | 0 | 0 | 1 |
| immusferr | 0 | 0 | 1 |
| syedhamzamohiuddin | 0 | 0 | 1 |
| beatsea20 | 0 | 0 | 1 |
| panhu | 0 | 0 | 1 |
| tirk999 | 0 | 0 | 1 |
| Non1187 | 0 | 0 | 1 |
| Sachi-27 | 0 | 0 | 1 |
| bossebandowski | 0 | 0 | 1 |
| SakshiFadnavis2003 | 0 | 0 | 1 |
| RaulCastillo547 | 0 | 0 | 1 |
| dwang6524 | 0 | 0 | 1 |
| kiriti-pendyala | 0 | 0 | 1 |
| durgas4 | 0 | 0 | 1 |
| buttaRahul | 0 | 0 | 1 |
| WojciechRynczuk | 0 | 0 | 1 |
| hotamago | 0 | 0 | 1 |
| AdityaB-01 | 0 | 0 | 1 |
| babvijayb | 0 | 0 | 1 |
| yayale1 | 0 | 0 | 1 |
| MohannadAbuIssa | 0 | 0 | 1 |
| Yulee3542 | 0 | 0 | 1 |
| a-sajjad72 | 0 | 0 | 1 |
| KeondoPark | 0 | 0 | 1 |
| coco-boy | 0 | 0 | 1 |
| dauso | 0 | 0 | 1 |
| BGigotSDS | 0 | 0 | 1 |
| risheek-mittal | 0 | 0 | 1 |
| eyalhir74 | 0 | 0 | 1 |
| hanan454 | 0 | 0 | 1 |
| wscJayasooriya | 0 | 0 | 1 |
| libofei2004 | 0 | 0 | 1 |
| maximd1 | 0 | 0 | 1 |
| 3CKTOR | 0 | 0 | 1 |
| dhruv2103 | 0 | 0 | 1 |
| nvn234 | 0 | 0 | 1 |
| AmarOk1412 | 0 | 0 | 1 |
| s3rv1c03t3rs3c | 0 | 0 | 1 |
| benjaminreynoso | 0 | 0 | 1 |
Don't miss what's next. Subscribe to Weekly Project News: