Migrate RL training recipes from verl to Relax frameworks.
npx skills add https://github.com/EnkeZ/tgsopd --skill verl-to-relax-enkez
Please help me install this Agent Skill. Skill: verl-to-relax Source: https://github.com/EnkeZ/tgsopd/tree/main/skills/verl-to-relax Command: npx skills add https://github.com/EnkeZ/tgsopd --skill verl-to-relax-enkez
This Skill simplifies the process of migrating reinforcement learning training recipes from the verl framework to the Relax framework, enabling users to seamlessly port their reward functions, tool environments, training scripts, and other recipe code.
Run the verl-to-relax skill with the following command: verl-to-relax --input-dir path/to/verl_recipe.
High-intent search queries and answers about installing and using this skill.
To migrate RL training recipes from verl to Relax, use the conversion utility that transforms reward functions, tool environments, training scripts, and dataset configurations. Run the command-line tool pointing to your verl recipe directory to start the migration.
Recipe migration from verl to Relax converts reward functions, tool environments, training scripts, and dataset configurations. This allows you to seamlessly port your existing reinforcement learning code to the new framework.
Yes, verl to Relax migration supports both colocate and fully async deployment modes. This provides flexibility when configuring how your reinforcement learning training scripts operate after porting your recipe code.
Yes, you need both the verl and Relax frameworks installed. The conversion process requires verl to read existing training recipes and Relax to deploy the migrated reward functions and environments.
The best way to convert verl dataset configurations and reward functions is to run the migration utility with your input directory specified. It handles the recipe conversion automatically, facilitating a seamless transition to Relax.
redai-infraConvert verl reinforcement learning recipes and code to Relax.
DeepExperienceConvert verl reinforcement learning recipes to Relax framework implementations.
DeepExperienceMigrates RedAccel reinforcement learning algorithms to Relax-compatible modules and interfaces.
hung-phanScale GRPO/PPO reinforcement learning training across multiple GPUs with DeepSpeed.
