Weekly GitHub Report for Tensorflow - 2024-07-29 12:00:57
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: 25
Summarized Issues:
- Inference Errors in Quantized TensorFlow Lite Models: Users encounter a ValueError when running inference on quantized TensorFlow Lite models, with the error message "Tensor data is null. Run allocate_tensors() first." This issue does not occur with non-quantized models, indicating a problem specific to quantization. The error suggests a need for additional steps or fixes in the quantization process.
- Tensor Dimension Discrepancies: There are issues related to tensor dimension mismatches in TensorFlow Lite models, particularly on Android devices. These mismatches result in runtime errors during operations like transpose. Such discrepancies highlight the need for better handling of tensor dimensions in mobile environments.
- Build Failures on ARM64 and RISC-V Platforms: Users report build failures when cross-compiling TensorFlow Lite with GPU support for ARM64-v8 Android devices and when compiling TensorFlow on RISC-V platforms. Errors are related to specific functions and missing members in the codebase, indicating compatibility issues with these architectures.
- Security Vulnerabilities: Several issues highlight security vulnerabilities in TensorFlow, such as CVE-2021-35958 and CVE-2023-33976. Users are concerned about the resolution and publication of these vulnerabilities, which have been flagged by security scans but are not yet addressed or reported on official platforms.
- API Crashes and Core Dumps: Issues with TensorFlow APIs like
tf.raw_ops.MapUnstageandtf.raw_ops.BlockLSTMV2result in crashes and core dumps. These crashes are due to tensor size mismatches and failed checks within the tensor framework, affecting multiple TensorFlow versions including nightly builds.
- Function Inconsistencies: Users report inconsistencies in TensorFlow functions like
tf.strided_sliceandtf.range. These issues involve unexpected behavior with specific parameters and unsupported data types, affecting operations like image processing and mixed precision in Keras layers.
- TensorFlow Lite on Microcontrollers: Running TensorFlow Lite models on microcontrollers like Cortex-M7 results in input and output tensors pointing to the same memory location. This causes the output to incorrectly reflect the input values, indicating a need for better memory management in such environments.
- Custom Metrics and Callbacks: Issues with custom metrics for SSIM and PSNR in TensorFlow models result in incorrect outputs when using the 'History' callback. The metrics return KerasVariable instances instead of numpy arrays, causing all training and validation values to equal the last printed validation value.
- Feature Requests and Documentation: Users request features like inspecting runtime-optimized
buffer_sizevalues withtf.data.AUTOTUNEand flagging incompatible tutorials with the latest stable release. These requests highlight gaps in current documentation and the need for better guidance on using new features.
- Distributed Training Issues: Problems with distributed training setups using
tf.distributeandkeras3 result in errors likeValueError: Invalid reduction dimension 0 for input with 0 dimensions. Users also face issues averaging gradients from multiple GPUs due to dimension mismatches, indicating challenges in distributed training configurations.
- Custom Code Execution in TensorFlow Serving: Users seek assistance on adding custom code to execute before or after specific GPU operations within TensorFlow Serving. They encounter difficulties in locating the appropriate function to modify, indicating a need for better documentation or support for such customizations.
- Native Method Loading Failures on Android: Issues with loading native TensorFlow Lite methods on Android devices result in
java.lang.UnsatisfiedLinkErrorandSecurityException. These errors are due to missing or improperly loaded native libraries and unavailable APIs, affecting the initialization of modules likeTfLiteVision.
- Compatibility and Module Errors: Users encounter errors like "DLL load failed while importing _pywrap_tensorflow_internal" on Python 3.12 and
ModuleNotFoundErrorfortensorflow.python.distribute.distribution_strategy_contextin TensorFlow 2.15. These issues indicate compatibility problems with specific Python versions and missing modules.
- OpKernel Registration Errors: Using
tf.distribute.MirroredStrategyon Windows 11 with TensorFlow 2.10.0 results inInvalidArgumentErrordue to multiple OpKernel registrations matching theAssignVariableOpNodeDef. This issue highlights challenges in kernel registration and priority handling in distributed strategies.
1.2 Top 5 Active Issues:
We consider active issues to be issues that have generated much discussion in the issue's comments.
- cuDNN, cuFFT, and cuBLAS Errors: This issue is about errors encountered when running TensorFlow on a GPU, specifically related to the inability to load cuDNN, cuFFT, and cuBLAS libraries, which results in slower training performance. The user is running TensorFlow on WSL2 with Ubuntu 22 and an NVIDIA Geforce GTX 1660 Ti GPU, and despite the GPU being recognized, the errors prevent optimal performance.
- 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. Users report mixed success, with some managing to get TensorFlow to recognize the GPU but still encountering the registration errors. The issue persists across different setups, including native Ubuntu installations, and there are suggestions that the problem might be related to the latest TensorFlow versions. Some users have found temporary workarounds by downgrading to TensorFlow 2.13 or using nightly builds, but a definitive solution is still pending.- Number of comments: 126
-
how to download or install .so file for tflite conversion with gpu delegate: This issue is about a user seeking guidance on how to download or install the
.sofile for TensorFlow Lite conversion with a GPU delegate. The user has provided detailed system information, code snippets, and specific questions about the conversion process and the errors encountered.- The comments section includes detailed troubleshooting steps, suggestions for building the
.sofile using Bazel, and discussions about compatibility issues with GPU delegates on different platforms. The user also shared their attempts and errors faced while trying to build and use the.sofile, and there were multiple exchanges seeking further clarification and assistance. - Number of comments: 55
- The comments section includes detailed troubleshooting steps, suggestions for building the
-
cuBLAS Error in 2.14.0: This issue reports a bug in TensorFlow version 2.14.0, where users encounter errors related to the registration of cuDNN, cuFFT, and cuBLAS factories when importing TensorFlow on various platforms, including Ubuntu and Windows with WSL2. The problem persists across different setups and configurations, causing significant inconvenience for users attempting to utilize GPU functionalities.
- Multiple users have confirmed experiencing the same issue across different operating systems and configurations, with some suggesting downgrading to earlier TensorFlow versions as a temporary workaround, while others report that the errors do not impact performance significantly.
- Number of comments: 53
-
TensorFlow Profiler did not work correctly.: This issue is about the TensorFlow Profiler not working correctly when measuring the performance of large-scale deep learning models on CPU/GPU, as the results are not displayed in TensorBoard. The user suspects that the current version of the TensorFlow Profiler does not save all the necessary files for measurement results.
- The comments discuss whether the issue is a bug, with some users confirming the same problem and others providing workarounds like manually copying files. There are suggestions to check the directory structure, TensorFlow, and TensorBoard versions, and some users offer to help debug and fix the issue. The conversation also includes references to related issues and the need for compatible versions of TensorFlow and TensorBoard plugins.
- Number of comments: 44
-
TF-TRT Warning: Could not find TensorRT: This issue is about a user experiencing difficulties with TensorFlow not being able to find TensorRT on their Ubuntu 22.04 system, despite multiple installation attempts and driver adjustments. The user is seeking help to resolve this problem as it is hindering their progress in a Machine Learning course.
- Multiple users report similar issues with TensorRT not being found by TensorFlow, despite various installation methods and environment setups. Suggestions include installing TensorRT separately, ensuring correct paths in the environment variables, and creating symbolic links for the TensorRT libraries. Some users found success by setting the
LD_LIBRARY_PATHcorrectly and creating symbolic links for the TensorRT libraries, while others still face issues despite these steps. - Number of comments: 41
- Multiple users report similar issues with TensorRT not being found by TensorFlow, despite various installation methods and environment setups. Suggestions include installing TensorRT separately, ensuring correct paths in the environment variables, and creating symbolic links for the TensorRT libraries. Some users found success by setting the
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.
-
Support for asynchronous execution in TensorFlow DLPack interface: This issue is about a request to add asynchronous execution support to the TensorFlow DLPack interface, specifically for the
tensorflow.experimental.dlpack.to_dlpackandtensorflow.experimental.dlpack.from_dlpackfunctions. The current synchronous operations can cause synchronization overhead when working across multiple frameworks, and the proposed asynchronous support could enhance execution efficiency and reduce this overhead.- Open for 366 days, 13 hours, 01 minutes
-
tf.debugging.experimental.enable_dump_debug_info (Debugger V2) error with TPU: This issue pertains to a bug encountered when using the
tf.debugging.experimental.enable_dump_debug_info(...)function in conjunction with a TPU in TensorFlow, specifically versions 2.12 and 2.14 nightly. The user reports that while the code works fine without the TPU strategy or the debugger individually, combining both results in a device assignment error, indicating an incompatibility between the TPU and the debugging function.- Open for 366 days, 10 hours, 29 minutes
-
Unexpected differences in outputs of Conv2D copy with exact subset of weights: This issue pertains to unexpected discrepancies in the outputs of a
Conv2Dlayer and its copy, even when the weights of certain channels are set to zero and the relevant parts of the weights are identical. The user observes differences in the outputs up to~1.7e-7and suspects that a different underlying mechanism might be causing these slight variations, which also occur in Dense layers.- Open for 366 days, 03 hours, 54 minutes
-
[TfLite] unresolved TfLiteGPUDelegateV2Create with Visual Studio: This issue pertains to a linker error encountered when attempting to build TensorFlow Lite with GPU support on Windows 10 using Visual Studio 2022. The error, specifically
LNK2001: unresolved external symbol __imp_TfLiteGpuDelegateV2Create, occurs despite the function being present in thetensorflow-lite.libfile, and the problem does not manifest when using the same build commands on Ubuntu Linux.- Open for 362 days, 04 hours, 53 minutes
-
Tflite use USB camera with android image classification app: This issue pertains to a user seeking assistance with modifying the TensorFlow Lite image classification app to utilize a USB camera instead of the default mobile back camera. The user is specifically asking for guidance on what changes need to be made in the
CameraFragment.ktfile to enable the app to detect and use a connected USB camera.- Open for 361 days, 20 hours, 41 minutes
1.4 Closed Issues
Closed Issues This Week: 20
Average Issue Close Time (This Week): 15.41 days
Summarized Issues:
- Buffer Overflow in TensorFlow Lite Kernel: This issue describes a potential buffer overflow in the
tensorflow/lite/kernels/stablehlo_pad.ccfile. The problem arises due to a size mismatch where pointers referencing 48-byte memory locations are passed to thememcpyfunction with a size parameter of 64 bytes. This could lead to memory corruption and unpredictable behavior.
- Lack of Non-Deprecated Alternatives in
tf.keras.preprocessing: This issue highlights the lack of non-deprecated alternatives in TensorFlow'stf.keras.preprocessingmodule. The deprecatedImageDataGeneratorlacks replacements for common image augmentations like random translations, rotations, and shear transformations. Users are left without modern tools to perform these essential tasks.
- Issues with TensorFlow Lite Conversion: Multiple issues describe problems with converting models to TensorFlow Lite format. One issue involves an LSTM-based Keras model failing to convert using
from_keras_modelbut working withfrom_saved_model. Another issue involves anAttributeErrorduring conversion, resolved by using the Legacy Keras workaround.
- Build and Installation Errors: Several issues report build and installation errors on different systems. Problems include build errors on Linux Ubuntu 22.04, difficulties integrating TensorFlow models into a website on Windows 11, and errors during TensorFlow Lite compilation for ARM64 architecture.
- AttributeError in TensorFlow Modules: Users have encountered
AttributeErrorissues in TensorFlow modules. One issue involves thetensorflow_model_optimizationmodule missing an attribute, while another involves an error in thetf.keras.layers.MaxPool1DAPI when using 'channels_first' data format.
- GPU Detection and Compatibility Issues: Issues have been reported regarding TensorFlow's inability to detect GPUs. One issue involves TensorFlow 2.14.0 failing to detect the GPU, while another seeks the recommended CUDA version for TensorFlow 2.17.0 on Ubuntu 22.04.
- Sparsemax Function Request: A user has requested the addition of a sparsemax function to TensorFlow. This function is necessary for TabNet and is difficult to implement independently. The request highlights the need for this feature in the TensorFlow ecosystem.
- Model Loading and Visualization Issues: Users have reported issues with loading models and visualizing training data. One issue involves a
ValueErrordue to missing model files, while another involves difficulties visualizing training data in TensorBoard.
- Errors Running Models and Tutorials: Several issues describe errors encountered while running models and tutorials. Problems include errors with the MoveNet model on Google Colab, access denied errors in the Fast Style Transfer tutorial, and
UnsatisfiedLinkErroron Android devices.
- Convolutional Autoencoder Output Issue: This issue describes a problem where updating TensorFlow from version 2.10 to 2.16.2 caused a convolutional autoencoder to output pure black images from the second epoch onward. The issue was resolved by changing the activation function from sigmoid to tanh.
- Miscellaneous Issues: Other issues include a spam report and an error in the "delicinalm" feature within the "perks" section of the TensorFlow project.
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.
-
cuDNN, cuFFT, and cuBLAS Errors
- Toxicity Score: 0.55 (Frustration, unresolved issues, repeated attempts, technical complexity)
- This GitHub conversation involves multiple users discussing issues related to TensorFlow, specifically around installation and configuration problems. The conversation starts with user Ke293-x2Ek-Qe-7-aE-B reporting errors when running TensorFlow with CUDA and cuDNN. SuryanarayanaY suggests a solution, which Ke293-x2Ek-Qe-7-aE-B tries but reports it still doesn't work. Other users, like AthiemoneZero and FaisalAlj, join in, sharing similar issues and their attempts to resolve them. The tone remains collaborative, with users sharing detailed steps and logs. However, frustration is evident as solutions are not consistently effective. The conversation includes technical discussions about environment setups, version compatibility, and specific error messages. Despite various suggestions and partial successes, the issue persists, leading to continued troubleshooting and sharing of experiences.
-
- Toxicity Score: 0.55 (Repeated frustration, unresolved issues, multiple users facing the same problem, direct language)
- This GitHub conversation involves multiple users, including Rajcr2, DineshNeupane, and devVegaAn, discussing a technical issue related to TensorFlow and Keras. The conversation starts with Rajcr2 reporting an error and seeking help. Various users provide potential solutions and workarounds, with some expressing frustration when the suggested fixes do not work. The tone remains mostly collaborative, although there are moments of tension, particularly when users repeatedly encounter the same issue despite trying different solutions. The conversation includes apologies for delays and requests for further clarification, indicating a generally helpful but occasionally strained interaction.
-
What is preventing TF to use GPU when used in native windows?
- Toxicity Score: 0.55 (Frustration, dissatisfaction, complexity of solutions, perceived neglect of Windows support)
- This GitHub conversation involves multiple users discussing the challenges and frustrations related to TensorFlow's GPU support on native Windows. User eabase initiates the conversation with a query, and several users, including mihaimaruseac and sgkouzias, provide insights and explanations. The tone of the conversation is generally constructive, though there are moments of frustration, particularly from eabase and other users who express dissatisfaction with the current state of support and the complexity of potential solutions. The conversation remains mostly respectful, but there are signs of tension, especially when discussing the perceived neglect of Windows support and the difficulties in compiling TensorFlow on Windows.
II. Pull Requests
2.1 Open Pull Requests
Open Pull Requests This Week: 1
Pull Requests:
- TensorFlow to TOSA operation legalization: This pull request focuses on adding support for legalizing the TensorFlow operation
scatter_ndto the TOSA operationscatter. It ensures that inputs are reformatted appropriately to match the TOSA operation requirements. Additionally, it restricts the legalization process to constant indices tensors with unique values.
2.2 Closed Pull Requests
Closed Pull Requests This Week: 10
Summarized Pull Requests:
- Quantization Support in TensorFlow: This topic covers the addition of per-axis and per-tensor quantization support to the
convolutionoperation in TensorFlow. The pull request also includes corresponding quantized unit tests to ensure the functionality works as expected. This enhancement aims to improve the flexibility and performance of TensorFlow's quantization capabilities.
- Buffer Overflow Fixes: This topic addresses a potential buffer overflow issue in the
tensorflow/lite/kernels/stablehlo_pad.ccfile. The fix involves changing thekMaxDimsvalue toTFLITE_STABLEHLO_PAD_PARAMS_MAX_DIMENSION_COUNT, as identified by the Svace static analyzer. This change aims to enhance the stability and security of TensorFlow Lite.
- Grammatical and Typographical Corrections: This topic involves making grammatical and typographical corrections to the file
tensorflow/tensorflow/lite/python/lite.pyin the TensorFlow project. The pull request also includes a link to the TFLiteConverter API documentation. These changes aim to improve the readability and accuracy of the documentation.
- Management of TF_CUDNN_DETERMINISTIC: This topic involves transferring the management of the
TF_CUDNN_DETERMINISTICenvironment variable from XLA to TensorFlow. The goal is to make the control of determinism behavior in XLA more explicit through numeric options. A related XLA pull request is expected to follow.
- Cherry-Picking Commits to r2.17 Branch: This topic covers cherry-picking specific commits into the r2.17 branch of the TensorFlow project. One commit includes necessary header (.h) files in the AAR library, while another fixes a typo in the GCS URI. These changes ensure that the r2.17 branch remains up-to-date and functional.
- Miscellaneous Pull Requests: This topic includes various other pull requests that do not fit into the above categories. One pull request, titled "Shoutout," is a non-technical submission described as "To the homies." Another pull request addresses missing datatype support in the TensorFlow Lite (TfLite) concat and pad operations, and another adds support for bf16 and f16 data types in TensorFlow Lite's min-max operations.
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: 226
Summarized Commits:
- Reverts and Fixes: Several commits involve reverting previous changes due to breakages or issues, such as reverting changes identified by hashes 91df7e393aa1361a215e7bc6e70480aff6111b41, 0af72d018f00cfefd7829bebc02482882ebea74a, 190932b47657829a9ec58d5c3f30de12aa7a2a56, and 1a98f7480924c696d76f182913e35c6b185f2be4. These reverts are necessary to maintain project stability and functionality.
- Relocations and Refactoring: Multiple commits involve relocating files and refactoring code for better organization and maintainability. Examples include moving
model_builderto the TFL compiler/converter directory, relocating helper functions to literal utils in XLA:GPU, and refactoring theEmitTargetElementLoopfunction within the XLA CPU backend.
- Debugging and Profiling Enhancements: Commits such as introducing
mutable_debug_options()toHloModuleConfig, retaining outer custom call names in the profiler for XLA:GPU, and adding annotations to instructions with their scheduling names in XLA:GPU enhance debugging and profiling capabilities.
- Data Race and Error Handling: Several commits address data race issues and improve error handling, such as resolving a data race in
coordination_service_test.cc, ensuringhlo_evaluatordoes not dereference a disengaged optional, and adding diagnostic information to error messages in the IFRT IR module.
- Memory and Performance Optimizations: Commits focus on optimizing memory usage and performance, such as removing redundant data transfers between host and device, optimizing the process of obtaining FLOPs for GPU instructions, and reserving elements in a hash set to improve performance in XLA:GPU.
- New Tests and Test Enhancements: Many commits introduce new tests or enhance existing ones, such as adding HLO-based pipeline parallelism tests, introducing tests for
dynamic-reshapeoperation in thunks runtime, and updating the base test case for circular pipelining to include a matmul operation.
- API and Version Updates: Commits include updates to APIs and versioning, such as introducing an API version to
XLA_FFI_Api, updating the GraphDef version to 1935, and updating the forward compatibility horizon to 2024-07-26.
- Build and Configuration Changes: Several commits involve changes to build rules and configurations, such as introducing build rules for proto pyclif, renaming environment variables associated with the installer wheel, and updating the build configuration for
//tensorflow/tools/lib_package:cheaders.
- New Features and Functionalities: Commits introduce new features and functionalities, such as support for Int16 activation and Int4 weight quantization, adding a physical device ordinal to run options, and introducing support for 3D convolution operations without padding.
- Code Simplification and Cleanup: Many commits focus on simplifying and cleaning up the codebase, such as removing unnecessary parentheses to resolve build breakage, cleaning up algebraic simplifier headers, and removing unused configuration settings like
aws_supportandhdfs_support.
- Integration and Compatibility: Commits involve integrating external projects and ensuring compatibility, such as integrating the LLVM project at specific commits, integrating the StableHLO component from the openxla/stablehlo repository, and updating the
@rules_pythondependency to version 0.34.0.
- Error and Bug Fixes: Numerous commits address and fix errors and bugs, such as fixing the test condition for the cuDNN workspace in the GPU module, addressing a linker error in
llvm_compiler_test, and resolving a thread sanitizer error inThunk::ExecuteState.
- Documentation and Comments: Commits enhance documentation and add comments for better code understanding, such as adding comments to the MsaAlgorithm class, enhancing comments and debugging logs in
HostOffloadLegalize, and improving documentation on XNNPack flag values.
- Runtime and Execution Improvements: Commits improve runtime and execution processes, such as introducing a non-blocking WhileThunk, ensuring donation transactions are committed even with thunk execution errors, and embedding LLVM IR into the executable for XLA:CPU backend.
- Quantization and Optimization: Commits focus on quantization and optimization techniques, such as introducing support for Int16 activation and Int4 weight quantization, optimizing the reshards to use a single RPC call, and enhancing the concatenate emitter by introducing vectorization for concats.
- Error Propagation and Handling: Commits introduce new methods for error propagation and handling, such as introducing long polling for error propagation in the coordination service and adding a runtime check to ensure
batch-norm-trainingoperation is rewritten before the emit phase.
- Thread and Resource Management: Commits address thread and resource management, such as creating an intra-op thread pool to prevent deadlocks, making the FusionInfoCache thread-safe, and enhancing the logging functionality for kShareable resource occupiers.
- Legalization and Conversion: Commits involve legalizing operations and converting between different formats, such as legalizing
mhlo.gatherto TensorFlow Lite, ensuring proper emission of float to int conversions, and introducing a pass to merge reshards with the same source and destination.
- Pipeline and Scheduling: Commits enhance pipeline and scheduling processes, such as refactoring post-scheduling passes for XLA:GPU, introducing a new pass to flatten tensors within XLA_GPU MLIR-based emitters, and updating the
scheduling_namefield inOpMetadata.
- Fusion and Autotuning: Commits focus on fusion and autotuning techniques, such as introducing a Custom Kernel Fusion Autotuner for XLA:GPU, adding a fusion pattern to combine redundant slice and pack operations, and optimizing the performance by computing tile offset indexing maps only when necessary.
- Simulation and Prediction: Commits introduce simulation and prediction tools, such as implementing a new simulator to predict execution time of HLO modules, introducing
SimulateComputeInstructionto simulate processing of asynchronous copy instructions, and adding a physical device ordinal to run options.
- Compatibility and Forward Horizon: Commits ensure forward compatibility and update horizons, such as updating the forward compatibility horizon to 2024-07-25, ensuring compatibility with future changes, and updating the GraphDef version to 1934.
- Automated Changes: Several commits involve automated code changes, as indicated by messages and PiperOrigin-RevId references, ensuring consistency and reducing manual intervention.
- Miscellaneous Enhancements: Various other enhancements include adding support for
transposeoperation in thunk runtime, introducing a unique fingerprint mechanism for XLA Expressions, and enhancing the robustness ofcpu_external_constants_test.
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 | 1170 | 0 | 0 |
| GitHub | 17 | 0 | 0 |
| tensorflow-jenkins | 0 | 7 | 0 |
| damjandakic93 | 0 | 0 | 3 |
| guillaume-tgl | 0 | 0 | 3 |
| libofei2004 | 0 | 0 | 3 |
| Ilia Sergachev | 2 | 0 | 0 |
| NobuoTsukamoto | 0 | 1 | 1 |
| lhutton1 | 0 | 2 | 0 |
| SsomsakTH | 0 | 1 | 1 |
| abhinav-mcw | 0 | 2 | 0 |
| Leo-Lifeblood | 0 | 0 | 2 |
| devapriyas2001 | 0 | 0 | 2 |
| eyalhir74 | 0 | 0 | 2 |
| hguandl | 0 | 0 | 2 |
| t-kalinowski | 0 | 0 | 2 |
| arthurflor23 | 0 | 0 | 2 |
| dryglicki | 0 | 0 | 2 |
| siy415 | 0 | 0 | 2 |
| KnightGOKU | 0 | 0 | 2 |
| A-fliga | 0 | 0 | 2 |
| 6eanut | 0 | 0 | 2 |
| pradeep10kumar | 0 | 0 | 2 |
| Brett Taylor | 1 | 0 | 0 |
| Pol Dellaiera | 1 | 0 | 0 |
| Yimei Sun | 1 | 0 | 0 |
| psunn | 0 | 1 | 0 |
| msteiner-google | 0 | 1 | 0 |
| yimeisun123 | 0 | 1 | 0 |
| drupol | 0 | 1 | 0 |
| vimalrajv | 0 | 1 | 0 |
| akhilgoe | 0 | 1 | 0 |
| RahulSundarMCW | 0 | 1 | 0 |
| alifbasha22 | 0 | 1 | 0 |
| tilakrayal | 0 | 1 | 0 |
| sergachev | 0 | 1 | 0 |
| gzmkl | 0 | 1 | 0 |
| yashpratap914 | 0 | 1 | 0 |
| quicbrtal | 0 | 1 | 0 |
| shadchin | 0 | 1 | 0 |
| Theorems-lab | 0 | 1 | 0 |
| eddie-santos | 0 | 1 | 0 |
| abhinavph21 | 0 | 1 | 0 |
| 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 |
| hanan454 | 0 | 0 | 1 |
| wscJayasooriya | 0 | 0 | 1 |
| maximd1 | 0 | 0 | 1 |
| dhruv2103 | 0 | 0 | 1 |
| nvn234 | 0 | 0 | 1 |
| AmarOk1412 | 0 | 0 | 1 |
| benjaminreynoso | 0 | 0 | 1 |
| Leli1024 | 0 | 0 | 1 |
| zhanghuicuc | 0 | 0 | 1 |
| prasad-mali07 | 0 | 0 | 1 |
| pulkitagarawal | 0 | 0 | 1 |
| Irayanbu05 | 0 | 0 | 1 |
| jiannanWang | 0 | 0 | 1 |
| Morehman27 | 0 | 0 | 1 |
| andy-tai | 0 | 0 | 1 |
| ewwll | 0 | 0 | 1 |
| Yongle-Fu | 0 | 0 | 1 |
| BREBION-Mathis | 0 | 0 | 1 |
| therooler | 0 | 0 | 1 |
| Ligeirinho00 | 0 | 0 | 1 |
| 333Random333 | 0 | 0 | 1 |
| SExpert12 | 0 | 0 | 1 |
| deepeshfujitsu | 0 | 0 | 1 |
| Tarun0000 | 0 | 0 | 1 |
| HashemZn-04 | 0 | 0 | 1 |
| Nayana-ibm | 0 | 0 | 1 |
| bernsny24 | 0 | 0 | 1 |
| 372046933 | 0 | 0 | 1 |
| tranvantungit | 0 | 0 | 1 |
| berndporr | 0 | 0 | 1 |
| Vinaygoudasp7 | 0 | 0 | 1 |
| jaskarannagi19 | 0 | 0 | 1 |
| VicB18 | 0 | 0 | 1 |
| 3cktorcrypto | 0 | 0 | 1 |
| Made-Jaya | 0 | 0 | 1 |
| Deepu777yt | 0 | 0 | 1 |
| poltomo | 0 | 0 | 1 |
| OlgasAcc | 0 | 0 | 1 |
| Allan1974 | 0 | 0 | 1 |
| arteen1000 | 0 | 0 | 1 |
| wonyoungmin | 0 | 0 | 1 |
| gajendrahatt | 0 | 0 | 1 |
| shkarupa-alex | 0 | 0 | 1 |
| JVD9kh96 | 0 | 0 | 1 |
| noahewolfe | 0 | 0 | 1 |
| justinvyu | 0 | 0 | 1 |
| mseDPYU4 | 0 | 0 | 1 |
| Jacob-yen | 0 | 0 | 1 |
| Andrew-XQY | 0 | 0 | 1 |
| SCH227 | 0 | 0 | 1 |
| Apprisco | 0 | 0 | 1 |
| Ayman250 | 0 | 0 | 1 |
| azhar47-sk | 0 | 0 | 1 |
| ThomasRichtsfeld | 0 | 0 | 1 |
| jmmelen | 0 | 0 | 1 |
Don't miss what's next. Subscribe to Weekly Project News: