Ready-to-run training recipes for reinforcement learning (GRPO, DAPO, GSPO, CISPO), preference optimization (DPO, ORPO), and supervised fine-tuning (SFT) on Fireworks.
개요
Ready-to-run training recipes for reinforcement learning (GRPO, DAPO, GSPO, CISPO), preference optimization (DPO, ORPO), and supervised fine-tuning (SFT) on Fireworks.
README
Fireworks AI Cookbook
Ready-to-run training recipes for reinforcement learning (GRPO, DAPO, GSPO, CISPO), preference optimization (DPO, ORPO), and supervised fine-tuning (SFT) on Fireworks.
Full documentation: Fireworks Training API
Quick Start
git clone https://github.com/fw-ai/cookbook.git
cd cookbook/training
conda create -n cookbook python=3.12 -y && conda activate cookbook
pip install --pre -e .
See training/README.md for configuration, recipes, and examples.
For AI Agents
One canonical skill covers the full training product: skills/fireworks-training/SKILL.md. It routes managed SFT/DPO/ORPO/RFT and Training API serverless or dedicated workflows, then progressively loads the relevant cookbook, operations, or troubleshooting references.
Claude Code
claude plugin marketplace add fw-ai/cookbook
claude plugin install fireworks-training@fw-ai-cookbook
Cursor
npx --yes skills add fw-ai/cookbook -g -s fireworks-training -a cursor -y
Codex
npx --yes skills add fw-ai/cookbook -g -s fireworks-training -a codex -y
The skill is portable Agent Skills Markdown. Cursor and Codex installation is
validated with the skills CLI; it is not limited to the Claude compact
interface. firectl may still require mutating commands to be run manually in
the user’s terminal when its AI-agent safety guard is active.
Repository Structure
training/ is the primary development surface. tools/ holds standalone
customer-facing scripts (public firectl workflows). Other top-level
directories (integrations/, multimedia/, archived/) are kept for backward
compatibility.
training/ Training API recipes, utilities, and examples
recipes/ Fork-and-customize training loop scripts
utils/ Shared config, data loading, losses, metrics
examples/ Worked examples (RL, SFT, DPO, ORPO)
verifier/ Renderer correctness validator + live React viewer
tests/ Unit and end-to-end tests
tools/ Standalone customer scripts (downloadable from GitHub)
skills/ Agent skills and reference docs
Training skill
skills/fireworks-training/SKILL.md— install once, then ask your coding agent to plan, run, resume, deploy, or debug training. The skill uses publicfirectlfor managed jobs and the cookbook for custom Training API loops.
Contributing
See the Contribution Guide.
Support
추천 도구
다른 키워드를 입력하거나 필터를 제거해 보세요.
설치
npx skillfish add fw-ai/cookbook