verl compatibility

2025-07-16

3 min read time

Applies to Linux and Windows

Volcano Engine Reinforcement Learning for LLMs (verl) is a reinforcement learning framework designed for large language models (LLMs). verl offers a scalable, open-source fine-tuning solution optimized for AMD Instinct GPUs with full ROCm support.

• See the verl documentation for more information about verl.

• The official verl GitHub repository is volcengine/verl.

• Use the AMD-validated Docker images with ROCm and verl preinstalled.

• See the ROCm verl installation guide to get started.

Note

verl is supported on ROCm 6.2.0.

Use cases and recommendations

The benefits of verl in large-scale reinforcement leaning from human feedback (RLHF) are discussed in the Reinforcement Learning from Human Feedback on AMD GPUs with verl and ROCm Integration blog.

Docker image compatibility

AMD validates and publishes ready-made ROCm verl Docker images with ROCm backends on Docker Hub. The following Docker image tags and associated inventories represent the latest verl version from the official Docker Hub. The Docker images have been validated for ROCm 6.2.0.

Docker image	verl	Linux	Pytorch	Python	vllm
rocm/verl	0.3.0post0	Ubuntu 20.04	2.5.0	3.9.19	0.6.4

Supported features

The following table shows verl and ROCm support for GPU-accelerated modules.

Module	Description	verl version	ROCm version
FSDP	Training engine	0.3.0.post0	6.2
vllm	Inference engine	0.3.0.post0	6.2