# MLOps & Training

> The engineering discipline around the model lifecycle: experiment tracking with MLflow and DVC, fine-tuning and RLHF with Axolotl, TRL, and unsloth, pipeline orchestration with Airflow and Kubeflow, model registries and feature stores, and core ML frameworks like PyTorch underneath it all.

45 tools in 5 subcategories. Web page: https://ailandscape.org/category/mlops-training · Each tool also has a markdown version at /tool/{slug}.md

## Experiment Tracking & CI/CD

- [CML](https://ailandscape.org/tool/cml): Continuous machine learning CI/CD tool
- [Comet ML](https://ailandscape.org/tool/comet-ml): ML experiment management platform
- [DVC](https://ailandscape.org/tool/dvc): Data version control for ML projects
- [GitHub Actions](https://ailandscape.org/tool/github-actions): Automate ML workflows with GitHub's CI/CD platform
- [GitLab CI](https://ailandscape.org/tool/gitlab-ci): Integrated CI/CD for ML pipelines in GitLab
- [MLflow](https://ailandscape.org/tool/mlflow): Open source platform for ML lifecycle
- [Neptune](https://ailandscape.org/tool/neptune): Experiment tracking for ML teams
- [Weights & Biases](https://ailandscape.org/tool/weights-biases): Developer tools for ML experiment tracking

## Fine-tuning & RLHF

- [Axolotl](https://ailandscape.org/tool/axolotl): Fine-tuning framework for LLMs supporting LoRA, QLoRA, DeepSpeed, and multi-GPU
- [OpenRLHF](https://ailandscape.org/tool/openrlhf): Scalable RLHF framework supporting 70B+ PPO full tuning, iterative DPO, and LoRA
- [TRL](https://ailandscape.org/tool/trl): Library for training transformer LMs with RLHF, SFT, reward modeling, and PPO
- [unsloth](https://ailandscape.org/tool/unsloth): 2-5x faster LLM fine-tuning with 70% less memory

## Pipeline Orchestration

- [Apache Airflow](https://ailandscape.org/tool/apache-airflow): Workflow automation and scheduling platform
- [Kubeflow](https://ailandscape.org/tool/kubeflow): Kubernetes-native ML pipeline orchestration
- [Prefect](https://ailandscape.org/tool/prefect): Modern workflow orchestration
- [ZenML](https://ailandscape.org/tool/zenml): MLOps framework for reproducible pipelines

## Model Registry & Feature Store

- [Feast](https://ailandscape.org/tool/feast): Open source feature store for ML
- [Hopsworks](https://ailandscape.org/tool/hopsworks): Feature store and MLOps platform
- [Hugging Face Hub](https://ailandscape.org/tool/hugging-face-hub): Community model repository and registry
- [MLflow Model Registry](https://ailandscape.org/tool/mlflow-model-registry): Centralized model store for lifecycle management
- [Tecton](https://ailandscape.org/tool/tecton): Enterprise feature platform for ML
- [Vertex Feature Store](https://ailandscape.org/tool/vertex-feature-store): Google Cloud's managed feature store for ML

## ML Frameworks

- [Accord.NET](https://ailandscape.org/tool/accord-net): .NET machine learning framework
- [Alink](https://ailandscape.org/tool/alink): Algorithm platform based on Flink by Alibaba
- [Angel-ML](https://ailandscape.org/tool/angel-ml): High-performance distributed machine learning platform
- [Auto-sklearn](https://ailandscape.org/tool/auto-sklearn): Automated machine learning toolkit
- [AutoGluon](https://ailandscape.org/tool/autogluon): AutoML toolkit by Amazon for text, image, and tabular data
- [Diffusers](https://ailandscape.org/tool/diffusers): State-of-the-art diffusion models library
- [FLAML](https://ailandscape.org/tool/flaml): Fast and Lightweight AutoML by Microsoft
- [Google AutoML](https://ailandscape.org/tool/google-automl): Google Cloud's automated ML service
- [H2O AutoML](https://ailandscape.org/tool/h2o-automl): Automatic machine learning from H2O.ai
- [Hugging Face Transformers](https://ailandscape.org/tool/hugging-face-transformers): State-of-the-art ML for PyTorch, TF, and JAX
- [JAX](https://ailandscape.org/tool/jax): Google's composable function transformations for ML
- [Keras](https://ailandscape.org/tool/keras): High-level neural networks API
- [LightGBM](https://ailandscape.org/tool/lightgbm): Fast gradient boosting framework by Microsoft
- [Mixture of Kittens](https://ailandscape.org/tool/mixture-of-kittens): Cursor's open-source MoE training megakernel for NVIDIA GB300 racks
- [OpenAI Gym](https://ailandscape.org/tool/openai-gym): Standard API for RL environments
- [OpenAI Python SDK](https://ailandscape.org/tool/openai-python-sdk): Official Python library for OpenAI API
- [PettingZoo](https://ailandscape.org/tool/pettingzoo): Multi-agent RL environment API
- [PyTorch](https://ailandscape.org/tool/pytorch): Open source deep learning framework by Meta
- [Ray RLlib](https://ailandscape.org/tool/ray-rllib): Scalable reinforcement learning library
- [scikit-learn](https://ailandscape.org/tool/scikit-learn): Machine learning in Python
- [Stable Baselines3](https://ailandscape.org/tool/stable-baselines3): Reliable RL algorithm implementations
- [TensorFlow](https://ailandscape.org/tool/tensorflow): Google's end-to-end open source ML platform
- [XGBoost](https://ailandscape.org/tool/xgboost): Optimized gradient boosting library

---

Part of [AI Landscape](https://ailandscape.org), an open map of the AI ecosystem. Index for AI assistants: https://ailandscape.org/llms.txt