The tooling section of KML.rocks is a collection of tools for doing machine learning on Kubernetes.
Kubeflow
Kubeflow is a Machine Learning Toolkit for Kubernetes, open sourced by Google in December 2017.
Getting started:
- Introducing Kubeflow - A Composable, Portable, Scalable ML Stack Built for Kubernetes
- Try out Kubeflow online for free (via Katacoda).
RADanalytics
The radanalytics.io project provides tooling with a focus on Apache Spark and Apache Kafka as well as tutorials covering examples from Tensorflow to Jupyter notebooks.
Machine Box
Machine Box is a collection of state of the art machine learning technologies inside a container. Currently the following boxes are available:
- facebox: detects and identifies faces in photos.
- textbox: natural language processing, with entity and keyword extraction, and sentiment analysis.
- tagbox: generate a list of descriptive tags for an image.
- nudebaox: detect images that contain nudity or adult content.
PaddlePaddle
PaddlePaddle is a Python-based deep learning platform provided by Baidu that can be run on Kubernetes.
Miscellaneous
- PipelineAI/pipeline—a standard runtime for real-time machine learning fully compatible with AWS SageMaker.
- SeldonIO/seldon-server—a machine learning platform, supports models built with TensorFlow, Keras, Vowpal Wabbit, XGBoost, Gensim.
- Open Neural Network Exchange to share/convert models between different tools.
- Langhalsdino/Kubernetes-GPU-Guide—a guide that explains how to automate and accelerate deep leaning training with Kubernetes; see noboevbo/machine-learning-kubernetes for a concrete example following this setup.