Grpo Python Packages

judgeval

The Continuous-Improvement Stack for Agents. Our environment data and evals power agent improvement and monitoring.

ms-swift

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-R1, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).

hud-python

OSS RL environment + evals toolkit

learnlens-rl

Universal evaluation layer for standard RL environments. Measures what an agent learned - not just how much reward it accumulated.

learnlens

Universal evaluation layer for standard RL environments. Measures what an agent learned - not just how much reward it accumulated.

lfm-coder

GRPO with RLVR training Liquid AI's LFM 2.5-instruct model to enhance coding capabilities

oat-llm

Online AlignmenT (OAT) for LLMs.

textpolicy

Reinforcement learning for text generation on MLX (Apple Silicon): GRPO/GSPO, environments, rollout, rewards, LoRA/QLoRA

r-torch

Open Implementation of Deepseek's R1

toolbrain

A framework for agentic tool use training with reinforcement learning

oxrl

A lightweight post-training framework for LLMs and VLMs

lightrft

LightRFT: Light, Efficient, Omni-modal & Reward-model Driven Reinforcement Fine-Tuning Framework

mlx-guided-grpo

Guided Group Relative Policy Optimization (GRPO) training for MLX on Apple Silicon

open-parl

PARL (Parallel-Agent Reinforcement Learning) is a training paradigm that teaches models to decompose complex tasks into parallel subtasks and coordinate multiple agents simultaneously.

genteki-hdp

OSS RL environment + evals toolkit