Skip to content

docs(deepeyes): remove missing run_deepeyes_4b.sh refs #15

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Yangruipis merged 1 commit into from
Apr 26, 2026
+6 −34
Merged

docs(deepeyes): remove missing run_deepeyes_4b.sh refs #15

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 3 additions & 17 deletions docs/en/examples/deepeyes.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,25 +49,12 @@ The HF Image dict format (`{"bytes": ...}`) is natively supported by Relax's ima
### Download the Model

```bash
hf download Qwen/Qwen3-VL-4B-Instruct \
--local-dir /root/Qwen3-VL-4B-Instruct
hf download Qwen/Qwen3-VL-30B-A3B-Thinking \
--local-dir /root/Qwen3-VL-30B-A3B-Thinking
```

For the full-scale configuration, use `Qwen/Qwen3-VL-30B-A3B-Thinking`.

## Quick Start

### 4B Model (8 GPUs)

```bash
export MODEL_DIR=/root
export DATA_DIR=/root
export SAVE_DIR=/root/save

cd /root/Relax
bash examples/deepeyes/run_deepeyes_4b.sh
```

### 30B-A3B Model (8 GPUs, MoE)

The full-scale configuration requires a judge model for reward scoring:
Expand All @@ -91,7 +78,7 @@ bash examples/deepeyes/run_deepeyes.sh
```bash
WORKING_DIR="./" RAY_ADDRESS=<RAY_HEAD_IP>:6379 \
MODEL_DIR=/root DATA_DIR=/root SAVE_DIR=/root/save \
bash -x scripts/entrypoint/ray-job.sh examples/deepeyes/run_deepeyes_4b.sh
bash -x scripts/entrypoint/ray-job.sh examples/deepeyes/run_deepeyes.sh
```

## Architecture
Expand All @@ -101,7 +88,6 @@ WORKING_DIR="./" RAY_ADDRESS=<RAY_HEAD_IP>:6379 \
```
examples/deepeyes/
├── run_deepeyes.sh # Launch script (Qwen3-VL-30B-A3B, full config)
├── run_deepeyes_4b.sh # Launch script (Qwen3-VL-4B, lightweight)
├── deepeyes_config.yaml # Task config (max_turns, env path)
├── rollout.py # Multi-turn rollout logic
├── env_deepeyes.py # DeepEyes tool-use environment
Expand Down
20 changes: 3 additions & 17 deletions docs/zh/examples/deepeyes.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -49,25 +49,12 @@ HF Image dict 格式（`{"bytes": ...}`）被 Relax 的图像加载管线原生
### 下载模型

```bash
hf download Qwen/Qwen3-VL-4B-Instruct \
--local-dir /root/Qwen3-VL-4B-Instruct
hf download Qwen/Qwen3-VL-30B-A3B-Thinking \
--local-dir /root/Qwen3-VL-30B-A3B-Thinking
```

完整配置使用 `Qwen/Qwen3-VL-30B-A3B-Thinking`。

## 快速开始

### 4B 模型（8 GPU）

```bash
export MODEL_DIR=/root
export DATA_DIR=/root
export SAVE_DIR=/root/save

cd /root/Relax
bash examples/deepeyes/run_deepeyes_4b.sh
```

### 30B-A3B 模型（8 GPU，MoE）

完整配置需要 judge 模型进行奖励评分：
Expand All @@ -91,7 +78,7 @@ bash examples/deepeyes/run_deepeyes.sh
```bash
WORKING_DIR="./" RAY_ADDRESS=<RAY_HEAD_IP>:6379 \
MODEL_DIR=/root DATA_DIR=/root SAVE_DIR=/root/save \
bash -x scripts/entrypoint/ray-job.sh examples/deepeyes/run_deepeyes_4b.sh
bash -x scripts/entrypoint/ray-job.sh examples/deepeyes/run_deepeyes.sh
```

## 架构
Expand All @@ -101,7 +88,6 @@ WORKING_DIR="./" RAY_ADDRESS=<RAY_HEAD_IP>:6379 \
```
examples/deepeyes/
├── run_deepeyes.sh # 启动脚本（Qwen3-VL-30B-A3B，完整配置）
├── run_deepeyes_4b.sh # 启动脚本（Qwen3-VL-4B，轻量配置）
├── deepeyes_config.yaml # 任务配置（max_turns、环境路径）
├── rollout.py # 多轮 rollout 逻辑
├── env_deepeyes.py # DeepEyes 工具使用环境
Expand Down
Loading