Examples

The examples/ directory contains a collection of self-contained examples that demonstrate how to use the TRL library for various applications. Each example lives in its own folder named after the method and the task it demonstrates (e.g. grpo_wordle, sft_gpt_oss), and holds everything the example needs: scripts, notebooks, prompts, chat templates, and evaluation code.

Basic single-trainer training scripts are not examples: they live in trl/scripts and are exposed through the command line interface (trl sft, trl dpo, trl grpo, …). Each trainer’s documentation page also contains a complete runnable snippet.

Shared resources sit at the root of examples/:

• examples/accelerate_configs: 🤗 Accelerate configuration files for multi-GPU, DeepSpeed ZeRO, FSDP, and context-parallel setups, used by many examples.
• examples/datasets: the scripts used to generate the trl-lib datasets used across the examples.

Getting Started

Install TRL and additional dependencies as follows:

pip install --upgrade trl[quantization]

Check for additional optional dependencies here. Notebook-based examples are self-contained and can run on free Colab; script-based examples run on single-GPU, multi-GPU, or DeepSpeed setups (see Distributed Training below).

Index

Example	Description	Open in Colab
async_distillation_math	Async on-policy distillation on GSM8K with experimental.async_distillation.AsyncDistillationTrainer: the teacher is served over HTTP with vLLM, including a multi-teacher (MOPD) math + code variant.
async_grpo_math	Asynchronous GRPO on GSM8K with experimental.async_grpo.AsyncGRPOTrainer, decoupling generation (vLLM server) from training.
async_grpo_opencode	AsyncGRPO training of the real opencode coding agent on an OpenEnv environment (loop-owning: the external agent runs its own tool loop and TRL trains on its captured proxy trace), with a local subprocess sandbox or remote Hugging Face sandboxes.
dpo_reduce_hallucinations	DPO fine-tuning of a Vision Language Model to reduce hallucinations using the openbmb/RLAIF-V-Dataset dataset.
gold_chatbot_arena	General Online Logit Distillation (GOLD) of a Qwen2 teacher into a Llama 3.2 student (cross-tokenizer) on chatbot_arena_completions with experimental.gold.GOLDTrainer, with full-training and LoRA variants.
gold_qwen3_vl	General Online Logit Distillation (GOLD) of Qwen3-VL-8B into smaller VLM students with experimental.gold.GOLDTrainer, covering same-family (JSD loss) and cross-family (ULD loss) distillation.
grpo_2048	GRPO with tool calling to teach a model to play the 2048 game.
grpo_browsergym	GRPO with the BrowserGym OpenEnv environment, with LLM and VLM variants.
grpo_carla	GRPO with the CARLA autonomous-driving OpenEnv environment, with LLM and VLM variants (multimodal camera-image tool responses).
grpo_catch	GRPO with the Catch (OpenSpiel) OpenEnv environment.
grpo_continuous_batching	GRPO with transformers’ continuous batching engine for faster generation on large batches with variable completion lengths.
grpo_echo	Minimal GRPO training with the Echo OpenEnv environment.
grpo_harbor	GRPO training against a Harbor task suite with a pluggable base agent (bash / jupyter / terminal_notes harnesses). See the Harbor Integration guide.
grpo_ministral3_vl	GRPO Ministral 3 with QLoRA on free Colab.
grpo_multi_env	Multi-environment GRPO training: Wordle + Catch OpenEnv environments in the same training run.
grpo_qlora	GRPO using QLoRA on free Colab.
grpo_qwen3_vl	GRPO Qwen3-VL with QLoRA on free Colab.
grpo_rnj_1_instruct	GRPO on rnj-1-instruct with QLoRA on Colab to add reasoning capabilities.
grpo_seta	GRPO training against the SETA ORS environment on the openreward.ai catalog. See the OpenReward Integration guide.
grpo_sql_agent	GRPO to train an agent that answers questions by querying a SQL database (script and notebook; not runnable on free Colab due to OOM).
grpo_sudoku	GRPO to play Sudoku on an OpenEnv environment (script and notebook).
grpo_visual_math	GRPO fine-tuning of a multimodal model for reasoning using the lmms-lab/multimodal-open-r1-8k-verified dataset.
grpo_wordle	GRPO to play Wordle (TextArena) on an OpenEnv environment (script and notebook).
gspo_math	GSPO via the GRPOTrainer for math reasoning on the AI-MO/NuminaMath-TIR dataset.
gspo_visual_math	GSPO via the GRPOTrainer to fine-tune a multimodal model for reasoning using the lmms-lab/multimodal-open-r1-8k-verified dataset.
mpo_visual_preferences	MPO via the DPOTrainer to align a multimodal model based on preferences using the HuggingFaceH4/rlaif-v_formatted dataset and a set of loss weights.
online_dpo_visual_math	Online DPO fine-tuning of a Vision Language Model with experimental.online_dpo.OnlineDPOTrainer.
ppo_sentiment	PPO with experimental.ppo.PPOTrainer to continue text with positive sentiment or physically descriptive language.
ppo_tldr	PPO with experimental.ppo.PPOTrainer to generate TL;DR summaries.
rloo_math	RLOO with the RLOOTrainer for math reasoning on the AI-MO/NuminaMath-TIR dataset with vLLM.
rloo_visual_math	RLOO fine-tuning of a multimodal model for reasoning using the lmms-lab/multimodal-open-r1-8k-verified dataset.
sdft_privileged_context	Self-distillation fine-tuning with experimental.sdft.SDFTTrainer, distilling privileged (teacher-only) context into the model.
sdpo_math	SDPO with experimental.sdpo.SDPOTrainer using verifiable math rewards and optional environment feedback on openai/gsm8k.
sft_diffusion_gemma	SFT of the DiffusionGemma block-diffusion language model on GSM8K by extending the SFTTrainer with a block-diffusion objective.
sft_gemma3	SFT of Gemma 3 on the Codeforces COTS dataset.
sft_gemma3_vision	SFT of Gemma 3 on vision to text tasks.
sft_gpt_oss	SFT of openai/gpt-oss-20b.
sft_ministral3_vl	SFT Ministral 3 with QLoRA on free Colab.
sft_nemotron_3	SFT of NVIDIA Nemotron 3 models (script and LoRA notebook).
sft_qlora	SFT using QLoRA on free Colab.
sft_qwen3_vl	SFT Qwen3-VL with QLoRA on free Colab.
sft_tool_calling	Teaching tool calling to a model without native tool-calling support using SFT with QLoRA (script, chat template, and notebook).
sft_visual_chat	SFT of a Vision Language Model in a chat setting. Only tested with LLaVA 1.5, LLaVA 1.6, and Llama-3.2-11B-Vision-Instruct; users may see unexpected behaviour in other model architectures.
ssd_codegen	Simple Self-Distillation for code generation with experimental.ssd.SSDTrainer, plus evaluation on LiveCodeBench.
tpo_ultrafeedback	Triple Preference Optimization with experimental.tpo.TPOTrainer using the tpo-alignment/triple-preference-ultrafeedback-40K dataset.

Distributed Training

You can run the example scripts on multiple GPUs with 🤗 Accelerate:

accelerate launch --config_file=examples/accelerate_configs/multi_gpu.yaml --num_processes {NUM_GPUS} path_to_script.py --all_arguments_of_the_script

For DeepSpeed ZeRO-{1,2,3}:

accelerate launch --config_file=examples/accelerate_configs/deepspeed_zero{1,2,3}.yaml --num_processes {NUM_GPUS} path_to_script.py --all_arguments_of_the_script

Adjust NUM_GPUS and --all_arguments_of_the_script as needed.