|
| 1 | +# SIG TF.js | |
| 2 | + |
| 3 | +## Objective |
| 4 | + |
| 5 | +To facilitate community-contributed components to tensorflow/tfjs. |
| 6 | + |
| 7 | +## Context |
| 8 | + |
| 9 | +JavaScript is a very versatile and widely-used language. With TensorFlow.js, web application |
| 10 | +developers and JS developers can now use Machine Learning with TensorFlow in their JS |
| 11 | +applications. TensorFlow.js is the leading framework for ML in JavaScript with huge user growth |
| 12 | +(2.1M npm downloads, more than 2x YoY growth in the number of avg daily downloads). With the |
| 13 | +growth of the library, we now support an increasing number of use cases across a variety of |
| 14 | +platforms and backends, with many more models, use-cases, and applications. |
| 15 | + |
| 16 | +The core TensorFlow.js engineering team has been working on building the infrastructure and |
| 17 | +tooling to enable ML to run in JavaScript powered applications, as well as adding support for |
| 18 | +more types of models. TensorFlow.js has more than 220 contributors on GitHub which include many |
| 19 | +passionate contributors (individual developers, GDEs, and enterprise users) who have been active |
| 20 | +participants in the TensorFlow.js ecosystem. We want to accelerate the community involvement in |
| 21 | +the project to help continue meet the needs and help drive new directions for the project. |
| 22 | + |
| 23 | +## Goals & Objectives |
| 24 | + |
| 25 | +We welcome community contributions on any area of TensorFlow.js, but the SIG will be focused on |
| 26 | +the following goals: |
| 27 | +- Driving TensorFlow.js features for Server-side execution with tfjs-node. This will include |
| 28 | +building tooling and infra for end to end, production-ready ML with TensorFlow.js. |
| 29 | +- Developing features for specific platforms and deployments (eg. React Native, IoT and embedded |
| 30 | +devices, etc); |
| 31 | +- Building workflows for Transfer learning features for customizing models |
| 32 | +- Adding support for newer models and application areas (eg. NLP, RL, etc), and data analysis |
| 33 | +algorithms |
| 34 | +- Conducting research and implementing prototypes for client-side, privacy preserving ML and |
| 35 | +Federated Learning. |
| 36 | +- Identifying performance issues, tradeoffs, and requirements. |
| 37 | +- Model benchmarking across browsers, hardware and devices. |
| 38 | +- Model encryption. |
| 39 | +- Model visualization. |
| 40 | +- These projects will move towards community ownership as work streams mature. |
| 41 | + |
| 42 | +## Membership |
| 43 | + |
| 44 | +We encourage any developers working at the intersection of machine learning and web/JS |
| 45 | +applications to join and participate in the activities of the SIG. Whether you are working on |
| 46 | +advancing the platform, prototyping or building specific applications, or authoring new libraries, |
| 47 | +we welcome your feedback on and contributions to Tensorflow.js and its tooling, and are eager to hear |
| 48 | +about any downstream results, implementations, and extensions. |
| 49 | + |
| 50 | +We have multiple channels for participation, and publicly archive discussions in our user group and |
| 51 | +announcements mailing lists: |
| 52 | +- [email protected] -- our general mailing list that all are welcome to join ( [archive group ](https://groups.google.com/a/tensorflow.org/g/tfjs)) |
| 53 | +- [email protected] -- Announcement only mailing list for TF.js community ( [archive group ](https://groups.google.com/a/tensorflow.org/g/tfjs-announce)) |
| 54 | + |
| 55 | +We will create a new mailing list for TensorFlow.js SIG members. |
| 56 | + |
| 57 | +### Other Resources |
| 58 | +- Repository: https://github.com/tensorflow/tfjs |
| 59 | +- Documentation: https://www.tensorflow.org/js |
| 60 | + |
| 61 | +## Organization and Governance |
| 62 | +SIG project leads will be added to a SIG TF.js maintainers team on GitHub to streamline their |
| 63 | +contributions to tensorflow/tfjs. As specific ideas and work streams develop, we will explore creating |
| 64 | +new, community-owned repos that the SIG will drive. |
| 65 | + |
| 66 | +## Contacts |
| 67 | +### Project Lead(s): |
| 68 | +- Ping Yu (Google) |
| 69 | +- Sandeep Gupta (Google) |
| 70 | +- Ton Ngo (IBM): |
| 71 | + - Server side support (GPU, training) |
| 72 | + - End-to-end pipeline on Kubernetes (Kubeflow) |
| 73 | +- Ricky Cao (Alibaba) |
| 74 | + - Building tooling and infra for end to end, production-ready ML with TensorFlow.js. |
| 75 | + - Pipeline with transfer learning |
| 76 | + - Model Encryption |
| 77 | +- Ningxin Hu (Intel) |
| 78 | + - Benchmarking and performance optimization |
| 79 | +- Gant Laborde (Infinite Red) |
| 80 | + - Platforms (React and React Native) |
| 81 | + - Adding support for newer models and application areas (eg. NLP, RL, etc), |
| 82 | + - Model visualization |
| 83 | + |
| 84 | +### SIG members: |
| 85 | +- Adam Bouhenguel (Tesserai) |
| 86 | +- Clément Walter (IBM) |
| 87 | +- Eric Li (Alibaba) |
| 88 | +- Mrinal Mathur (ARM) |
| 89 | +- Patrick Haralabidis (Carlisle Homes) |
| 90 | +- Paul Van Eck (IBM) |
| 91 | +- Rising Odegua (Datopian) |
| 92 | +- Saswat Samal (Student) |
| 93 | +- Shivay Lamba (Metvy) |
| 94 | +- Ted Chang (IBM) |
| 95 | +- Va Barbosa (IBM) |
| 96 | +- YiHong Wang (IBM) |
| 97 | +- Zebai (Alibaba) |
0 commit comments