alignment-handbook
Robust recipes to align language models with human and AI preferences
Stars: 4548
The Alignment Handbook provides robust training recipes for continuing pretraining and aligning language models with human and AI preferences. It includes techniques such as continued pretraining, supervised fine-tuning, reward modeling, rejection sampling, and direct preference optimization (DPO). The handbook aims to fill the gap in public resources on training these models, collecting data, and measuring metrics for optimal downstream performance.
README:
🤗 Models & Datasets | 📃 Technical Report
Robust recipes to continue pretraining and to align language models with human and AI preferences.
Just one year ago, chatbots were out of fashion and most people hadn't heard about techniques like Reinforcement Learning from Human Feedback (RLHF) to align language models with human preferences. Then, OpenAI broke the internet with ChatGPT and Meta followed suit by releasing the Llama series of language models which enabled the ML community to build their very own capable chatbots. This has led to a rich ecosystem of datasets and models that have mostly focused on teaching language models to follow instructions through supervised fine-tuning (SFT).
However, we know from the InstructGPT and Llama2 papers that significant gains in helpfulness and safety can be had by augmenting SFT with human (or AI) preferences. At the same time, aligning language models to a set of preferences is a fairly novel idea and there are few public resources available on how to train these models, what data to collect, and what metrics to measure for best downstream performance.
The Alignment Handbook aims to fill that gap by providing the community with a series of robust training recipes that span the whole pipeline.
- August 18, 2024: We release SmolLM-Instruct v0.2, along with the recipe to fine-tuning small LLMs 💻
- April 12, 2024: We release Zephyr 141B (A35B), in collaboration with Argilla and Kaist AI, along with the recipe to fine-tune Mixtral 8x22B with ORPO 🪁
- March 12, 2024: We release StarChat2 15B, along with the recipe to train capable coding assistants 🌟
- March 1, 2024: We release Zephyr 7B Gemma, which is a new recipe to align Gemma 7B with RLAIF 🔥
- February 1, 2024: We release a recipe to align open LLMs with Constitutional AI 📜! See the recipe and the blog post for details.
- January 18, 2024: We release a suite of evaluations of DPO vs KTO vs IPO, see the recipe and the blog post for details.
- November 10, 2023: We release all the training code to replicate Zephyr-7b-β 🪁! We also release No Robots, a brand new dataset of 10,000 instructions and demonstrations written entirely by skilled human annotators.
This project is simple by design and mostly consists of:
-
scripts
to train and evaluate models. Four steps are included: continued pretraining, supervised-finetuning (SFT) for chat, preference alignment with DPO, and supervised-finetuning with preference alignment with ORPO. Each script supports distributed training of the full model weights with DeepSpeed ZeRO-3, or LoRA/QLoRA for parameter-efficient fine-tuning. -
recipes
to reproduce models like Zephyr 7B. Each recipe takes the form of a YAML file which contains all the parameters associated with a single training run. Agpt2-nl
recipe is also given to illustrate how this handbook can be used for language or domain adaptation, e.g. by continuing to pretrain on a different language, and then SFT and DPO tuning the result.
We are also working on a series of guides to explain how methods like direct preference optimization (DPO) work, along with lessons learned from gathering human preferences in practice. To get started, we recommend the following:
- Follow the installation instructions to set up your environment etc.
- Replicate Zephyr-7b-β by following the recipe instructions.
If you would like to train chat models on your own datasets, we recommend following the dataset formatting instructions here.
The initial release of the handbook will focus on the following techniques:
- Continued pretraining: adapt language models to a new language or domain, or simply improve it by continued pretraining (causal language modeling) on a new dataset.
- Supervised fine-tuning: teach language models to follow instructions and tips on how to collect and curate your training dataset.
- Reward modeling: teach language models to distinguish model responses according to human or AI preferences.
- Rejection sampling: a simple, but powerful technique to boost the performance of your SFT model.
- Direct preference optimisation (DPO): a powerful and promising alternative to PPO.
- Odds Ratio Preference Optimisation (ORPO): a technique to fine-tune language models with human preferences, combining SFT and DPO in a single stage.
To run the code in this project, first, create a Python virtual environment using e.g. Conda:
conda create -n handbook python=3.10 && conda activate handbook
Next, install PyTorch v2.1.2
- the precise version is important for reproducibility! Since this is hardware-dependent, we
direct you to the PyTorch Installation Page.
You can then install the remaining package dependencies as follows:
git clone https://github.com/huggingface/alignment-handbook.git
cd ./alignment-handbook/
python -m pip install .
You will also need Flash Attention 2 installed, which can be done by running:
python -m pip install flash-attn --no-build-isolation
Note If your machine has less than 96GB of RAM and many CPU cores, reduce the
MAX_JOBS
arguments, e.g.MAX_JOBS=4 pip install flash-attn --no-build-isolation
Next, log into your Hugging Face account as follows:
huggingface-cli login
Finally, install Git LFS so that you can push models to the Hugging Face Hub:
sudo apt-get install git-lfs
You can now check out the scripts
and recipes
directories for instructions on how to train some models 🪁!
├── LICENSE
├── Makefile <- Makefile with commands like `make style`
├── README.md <- The top-level README for developers using this project
├── chapters <- Educational content to render on hf.co/learn
├── recipes <- Recipe configs, accelerate configs, slurm scripts
├── scripts <- Scripts to train and evaluate chat models
├── setup.cfg <- Installation config (mostly used for configuring code quality & tests)
├── setup.py <- Makes project pip installable (pip install -e .) so `alignment` can be imported
├── src <- Source code for use in this project
└── tests <- Unit tests
If you find the content of this repo useful in your work, please cite it as follows via \usepackage{biblatex}
:
@software{Tunstall_The_Alignment_Handbook,
author = {Tunstall, Lewis and Beeching, Edward and Lambert, Nathan and Rajani, Nazneen and Huang, Shengyi and Rasul, Kashif and Bartolome, Alvaro and M. Rush, Alexander and Wolf, Thomas},
license = {Apache-2.0},
title = {{The Alignment Handbook}},
url = {https://github.com/huggingface/alignment-handbook},
version = {0.3.0.dev0}
}
For Tasks:
Click tags to check more tools for each tasksFor Jobs:
Alternative AI tools for alignment-handbook
Similar Open Source Tools
alignment-handbook
The Alignment Handbook provides robust training recipes for continuing pretraining and aligning language models with human and AI preferences. It includes techniques such as continued pretraining, supervised fine-tuning, reward modeling, rejection sampling, and direct preference optimization (DPO). The handbook aims to fill the gap in public resources on training these models, collecting data, and measuring metrics for optimal downstream performance.
ChainForge
ChainForge is a visual programming environment for battle-testing prompts to LLMs. It is geared towards early-stage, quick-and-dirty exploration of prompts, chat responses, and response quality that goes beyond ad-hoc chatting with individual LLMs. With ChainForge, you can: * Query multiple LLMs at once to test prompt ideas and variations quickly and effectively. * Compare response quality across prompt permutations, across models, and across model settings to choose the best prompt and model for your use case. * Setup evaluation metrics (scoring function) and immediately visualize results across prompts, prompt parameters, models, and model settings. * Hold multiple conversations at once across template parameters and chat models. Template not just prompts, but follow-up chat messages, and inspect and evaluate outputs at each turn of a chat conversation. ChainForge comes with a number of example evaluation flows to give you a sense of what's possible, including 188 example flows generated from benchmarks in OpenAI evals. This is an open beta of Chainforge. We support model providers OpenAI, HuggingFace, Anthropic, Google PaLM2, Azure OpenAI endpoints, and Dalai-hosted models Alpaca and Llama. You can change the exact model and individual model settings. Visualization nodes support numeric and boolean evaluation metrics. ChainForge is built on ReactFlow and Flask.
aici
The Artificial Intelligence Controller Interface (AICI) lets you build Controllers that constrain and direct output of a Large Language Model (LLM) in real time. Controllers are flexible programs capable of implementing constrained decoding, dynamic editing of prompts and generated text, and coordinating execution across multiple, parallel generations. Controllers incorporate custom logic during the token-by-token decoding and maintain state during an LLM request. This allows diverse Controller strategies, from programmatic or query-based decoding to multi-agent conversations to execute efficiently in tight integration with the LLM itself.
feedgen
FeedGen is an open-source tool that uses Google Cloud's state-of-the-art Large Language Models (LLMs) to improve product titles, generate more comprehensive descriptions, and fill missing attributes in product feeds. It helps merchants and advertisers surface and fix quality issues in their feeds using Generative AI in a simple and configurable way. The tool relies on GCP's Vertex AI API to provide both zero-shot and few-shot inference capabilities on GCP's foundational LLMs. With few-shot prompting, users can customize the model's responses towards their own data, achieving higher quality and more consistent output. FeedGen is an Apps Script based application that runs as an HTML sidebar in Google Sheets, allowing users to optimize their feeds with ease.
bpf-developer-tutorial
This is a development tutorial for eBPF based on CO-RE (Compile Once, Run Everywhere). It provides practical eBPF development practices from beginner to advanced, including basic concepts, code examples, and real-world applications. The tutorial focuses on eBPF examples in observability, networking, security, and more. It aims to help eBPF application developers quickly grasp eBPF development methods and techniques through examples in languages such as C, Go, and Rust. The tutorial is structured with independent eBPF tool examples in each directory, covering topics like kprobes, fentry, opensnoop, uprobe, sigsnoop, execsnoop, exitsnoop, runqlat, hardirqs, and more. The project is based on libbpf and frameworks like libbpf, Cilium, libbpf-rs, and eunomia-bpf for development.
EdgeChains
EdgeChains is an open-source chain-of-thought engineering framework tailored for Large Language Models (LLMs)- like OpenAI GPT, LLama2, Falcon, etc. - With a focus on enterprise-grade deployability and scalability. EdgeChains is specifically designed to **orchestrate** such applications. At EdgeChains, we take a unique approach to Generative AI - we think Generative AI is a deployment and configuration management challenge rather than a UI and library design pattern challenge. We build on top of a tech that has solved this problem in a different domain - Kubernetes Config Management - and bring that to Generative AI. Edgechains is built on top of jsonnet, originally built by Google based on their experience managing a vast amount of configuration code in the Borg infrastructure.
CoLLM
CoLLM is a novel method that integrates collaborative information into Large Language Models (LLMs) for recommendation. It converts recommendation data into language prompts, encodes them with both textual and collaborative information, and uses a two-step tuning method to train the model. The method incorporates user/item ID fields in prompts and employs a conventional collaborative model to generate user/item representations. CoLLM is built upon MiniGPT-4 and utilizes pretrained Vicuna weights for training.
parlant
Parlant is a structured approach to building and guiding customer-facing AI agents. It allows developers to create and manage robust AI agents, providing specific feedback on agent behavior and helping understand user intentions better. With features like guidelines, glossary, coherence checks, dynamic context, and guided tool use, Parlant offers control over agent responses and behavior. Developer-friendly aspects include instant changes, Git integration, clean architecture, and type safety. It enables confident deployment with scalability, effective debugging, and validation before deployment. Parlant works with major LLM providers and offers client SDKs for Python and TypeScript. The tool facilitates natural customer interactions through asynchronous communication and provides a chat UI for testing new behaviors before deployment.
ask-astro
Ask Astro is an open-source reference implementation of Andreessen Horowitz's LLM Application Architecture built by Astronomer. It provides an end-to-end example of a Q&A LLM application used to answer questions about Apache Airflow® and Astronomer. Ask Astro includes Airflow DAGs for data ingestion, an API for business logic, a Slack bot, a public UI, and DAGs for processing user feedback. The tool is divided into data retrieval & embedding, prompt orchestration, and feedback loops.
airflow
Apache Airflow (or simply Airflow) is a platform to programmatically author, schedule, and monitor workflows. When workflows are defined as code, they become more maintainable, versionable, testable, and collaborative. Use Airflow to author workflows as directed acyclic graphs (DAGs) of tasks. The Airflow scheduler executes your tasks on an array of workers while following the specified dependencies. Rich command line utilities make performing complex surgeries on DAGs a snap. The rich user interface makes it easy to visualize pipelines running in production, monitor progress, and troubleshoot issues when needed.
gen-cv
This repository is a rich resource offering examples of synthetic image generation, manipulation, and reasoning using Azure Machine Learning, Computer Vision, OpenAI, and open-source frameworks like Stable Diffusion. It provides practical insights into image processing applications, including content generation, video analysis, avatar creation, and image manipulation with various tools and APIs.
TinyTroupe
TinyTroupe is an experimental Python library that leverages Large Language Models (LLMs) to simulate artificial agents called TinyPersons with specific personalities, interests, and goals in simulated environments. The focus is on understanding human behavior through convincing interactions and customizable personas for various applications like advertisement evaluation, software testing, data generation, project management, and brainstorming. The tool aims to enhance human imagination and provide insights for better decision-making in business and productivity scenarios.
llama-github
Llama-github is a powerful tool that helps retrieve relevant code snippets, issues, and repository information from GitHub based on queries. It empowers AI agents and developers to solve coding tasks efficiently. With features like intelligent GitHub retrieval, repository pool caching, LLM-powered question analysis, and comprehensive context generation, llama-github excels at providing valuable knowledge context for development needs. It supports asynchronous processing, flexible LLM integration, robust authentication options, and logging/error handling for smooth operations and troubleshooting. The vision is to seamlessly integrate with GitHub for AI-driven development solutions, while the roadmap focuses on empowering LLMs to automatically resolve complex coding tasks.
codebase-context-spec
The Codebase Context Specification (CCS) project aims to standardize embedding contextual information within codebases to enhance understanding for both AI and human developers. It introduces a convention similar to `.env` and `.editorconfig` files but focused on documenting code for both AI and humans. By providing structured contextual metadata, collaborative documentation guidelines, and standardized context files, developers can improve code comprehension, collaboration, and development efficiency. The project includes a linter for validating context files and provides guidelines for using the specification with AI assistants. Tooling recommendations suggest creating memory systems, IDE plugins, AI model integrations, and agents for context creation and utilization. Future directions include integration with existing documentation systems, dynamic context generation, and support for explicit context overriding.
pathway
Pathway is a Python data processing framework for analytics and AI pipelines over data streams. It's the ideal solution for real-time processing use cases like streaming ETL or RAG pipelines for unstructured data. Pathway comes with an **easy-to-use Python API** , allowing you to seamlessly integrate your favorite Python ML libraries. Pathway code is versatile and robust: **you can use it in both development and production environments, handling both batch and streaming data effectively**. The same code can be used for local development, CI/CD tests, running batch jobs, handling stream replays, and processing data streams. Pathway is powered by a **scalable Rust engine** based on Differential Dataflow and performs incremental computation. Your Pathway code, despite being written in Python, is run by the Rust engine, enabling multithreading, multiprocessing, and distributed computations. All the pipeline is kept in memory and can be easily deployed with **Docker and Kubernetes**. You can install Pathway with pip: `pip install -U pathway` For any questions, you will find the community and team behind the project on Discord.
aigt
AIGT is a repository containing scripts for deep learning in guided medical interventions, focusing on ultrasound imaging. It provides a complete workflow from formatting and annotations to real-time model deployment. Users can set up an Anaconda environment, run Slicer notebooks, acquire tracked ultrasound data, and process exported data for training. The repository includes tools for segmentation, image export, and annotation creation.
For similar tasks
alignment-handbook
The Alignment Handbook provides robust training recipes for continuing pretraining and aligning language models with human and AI preferences. It includes techniques such as continued pretraining, supervised fine-tuning, reward modeling, rejection sampling, and direct preference optimization (DPO). The handbook aims to fill the gap in public resources on training these models, collecting data, and measuring metrics for optimal downstream performance.
Xwin-LM
Xwin-LM is a powerful and stable open-source tool for aligning large language models, offering various alignment technologies like supervised fine-tuning, reward models, reject sampling, and reinforcement learning from human feedback. It has achieved top rankings in benchmarks like AlpacaEval and surpassed GPT-4. The tool is continuously updated with new models and features.
Awesome-LLM-Preference-Learning
The repository 'Awesome-LLM-Preference-Learning' is the official repository of a survey paper titled 'Towards a Unified View of Preference Learning for Large Language Models: A Survey'. It contains a curated list of papers related to preference learning for Large Language Models (LLMs). The repository covers various aspects of preference learning, including on-policy and off-policy methods, feedback mechanisms, reward models, algorithms, evaluation techniques, and more. The papers included in the repository explore different approaches to aligning LLMs with human preferences, improving mathematical reasoning in LLMs, enhancing code generation, and optimizing language model performance.
LLM-Synthetic-Data
LLM-Synthetic-Data is a repository focused on real-time, fine-grained LLM-Synthetic-Data generation. It includes methods, surveys, and application areas related to synthetic data for language models. The repository covers topics like pre-training, instruction tuning, model collapse, LLM benchmarking, evaluation, and distillation. It also explores application areas such as mathematical reasoning, code generation, text-to-SQL, alignment, reward modeling, long context, weak-to-strong generalization, agent and tool use, vision and language, factuality, federated learning, generative design, and safety.
athina-evals
Athina is an open-source library designed to help engineers improve the reliability and performance of Large Language Models (LLMs) through eval-driven development. It offers plug-and-play preset evals for catching and preventing bad outputs, measuring model performance, running experiments, A/B testing models, detecting regressions, and monitoring production data. Athina provides a solution to the flaws in current LLM developer workflows by offering rapid experimentation, customizable evaluators, integrated dashboard, consistent metrics, historical record tracking, and easy setup. It includes preset evaluators for RAG applications and summarization accuracy, as well as the ability to write custom evals. Athina's evals can run on both development and production environments, providing consistent metrics and removing the need for manual infrastructure setup.
agentcloud
AgentCloud is an open-source platform that enables companies to build and deploy private LLM chat apps, empowering teams to securely interact with their data. It comprises three main components: Agent Backend, Webapp, and Vector Proxy. To run this project locally, clone the repository, install Docker, and start the services. The project is licensed under the GNU Affero General Public License, version 3 only. Contributions and feedback are welcome from the community.
chatbot-ui
Chatbot UI is an open-source AI chat app that allows users to create and deploy their own AI chatbots. It is easy to use and can be customized to fit any need. Chatbot UI is perfect for businesses, developers, and anyone who wants to create a chatbot.
botpress
Botpress is a platform for building next-generation chatbots and assistants powered by OpenAI. It provides a range of tools and integrations to help developers quickly and easily create and deploy chatbots for various use cases.
For similar jobs
alignment-handbook
The Alignment Handbook provides robust training recipes for continuing pretraining and aligning language models with human and AI preferences. It includes techniques such as continued pretraining, supervised fine-tuning, reward modeling, rejection sampling, and direct preference optimization (DPO). The handbook aims to fill the gap in public resources on training these models, collecting data, and measuring metrics for optimal downstream performance.
ChatFAQ
ChatFAQ is an open-source comprehensive platform for creating a wide variety of chatbots: generic ones, business-trained, or even capable of redirecting requests to human operators. It includes a specialized NLP/NLG engine based on a RAG architecture and customized chat widgets, ensuring a tailored experience for users and avoiding vendor lock-in.
agentcloud
AgentCloud is an open-source platform that enables companies to build and deploy private LLM chat apps, empowering teams to securely interact with their data. It comprises three main components: Agent Backend, Webapp, and Vector Proxy. To run this project locally, clone the repository, install Docker, and start the services. The project is licensed under the GNU Affero General Public License, version 3 only. Contributions and feedback are welcome from the community.
anything-llm
AnythingLLM is a full-stack application that enables you to turn any document, resource, or piece of content into context that any LLM can use as references during chatting. This application allows you to pick and choose which LLM or Vector Database you want to use as well as supporting multi-user management and permissions.
ai-guide
This guide is dedicated to Large Language Models (LLMs) that you can run on your home computer. It assumes your PC is a lower-end, non-gaming setup.
Magick
Magick is a groundbreaking visual AIDE (Artificial Intelligence Development Environment) for no-code data pipelines and multimodal agents. Magick can connect to other services and comes with nodes and templates well-suited for intelligent agents, chatbots, complex reasoning systems and realistic characters.
glide
Glide is a cloud-native LLM gateway that provides a unified REST API for accessing various large language models (LLMs) from different providers. It handles LLMOps tasks such as model failover, caching, key management, and more, making it easy to integrate LLMs into applications. Glide supports popular LLM providers like OpenAI, Anthropic, Azure OpenAI, AWS Bedrock (Titan), Cohere, Google Gemini, OctoML, and Ollama. It offers high availability, performance, and observability, and provides SDKs for Python and NodeJS to simplify integration.
chatbot-ui
Chatbot UI is an open-source AI chat app that allows users to create and deploy their own AI chatbots. It is easy to use and can be customized to fit any need. Chatbot UI is perfect for businesses, developers, and anyone who wants to create a chatbot.