https://github.com/dvampire/verl_vlm
https://github.com/dvampire/verl_vlm
Last synced: about 1 year ago
JSON representation
- Host: GitHub
- URL: https://github.com/dvampire/verl_vlm
- Owner: DVampire
- License: apache-2.0
- Created: 2025-02-24T11:50:39.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-02-25T11:15:51.000Z (about 1 year ago)
- Last Synced: 2025-02-25T12:22:49.378Z (about 1 year ago)
- Language: Python
- Size: 1.72 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# VERL_VLM
The vlm version of verl, which currently only supports Qwen 2.5-VL.
# Installation
```bash
# Create the conda environment
conda create -n verl python==3.10
conda activate verl
pip3 install -e .
# Install vLLM>=0.7
pip3 uninstall vllm
pip3 install vllm==0.7.3 # must be v0.7.3
# Install flash-attn
pip3 install --use-pep517 flash-attn --no-build-isolation
pip3 uninstall torch
# Support CUDA 12.1
pip3 install torch==2.5.1 torchvision==0.20.1 torchaudio==2.5.1 --index-url https://download.pytorch.org/whl/cu121
pip3 install qwen_vl_utils
```
# RUN
```bash
python prepare_data.py
sh run_qwen2.5-vl-3b.sh
```