Hugging Face's logo

Datasets:
DesonDai
/
UIPress

Dataset card Files
xet
 Community
UIPress
File size: 3,599 Bytes
bb01096
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
#!/bin/bash
# =============================================================
# UIPress: 环境搭建脚本 (双环境)
#
# DeepSeek-OCR 需要 transformers==4.46.3
# Qwen VL 需要 transformers>=4.51
# 两者不兼容，因此需要两个独立的 conda 环境。
#
# 用法:
#   bash scripts/setup_env.sh          # 搭建两个环境
#   bash scripts/setup_env.sh deepseek # 只搭建 DeepSeek 环境
#   bash scripts/setup_env.sh qwen     # 只搭建 Qwen 环境
# =============================================================

set -e
PROJ_DIR="$(cd "$(dirname "$0")/.." && pwd)"
cd "$PROJ_DIR"

# 国内镜像
export HF_ENDPOINT=${HF_ENDPOINT:-https://hf-mirror.com}
export HF_HOME=${HF_HOME:-/root/rivermind-data/huggingface}
mkdir -p "$HF_HOME"

TARGET=${1:-all}

# --- 检测 CUDA 版本 ---
detect_cuda() {
    if command -v nvcc &>/dev/null; then
        CUDA_VER=$(nvcc --version | grep -oP 'release \K[0-9]+\.[0-9]+')
    elif [ -f /usr/local/cuda/version.txt ]; then
        CUDA_VER=$(cat /usr/local/cuda/version.txt | grep -oP '[0-9]+\.[0-9]+')
    else
        CUDA_VER="11.8"
    fi

    if [[ "$CUDA_VER" == 12.* ]]; then
        TORCH_INDEX="https://download.pytorch.org/whl/cu121"
    else
        TORCH_INDEX="https://download.pytorch.org/whl/cu118"
    fi
    echo "  Detected CUDA $CUDA_VER -> PyTorch index: $TORCH_INDEX"
}

setup_deepseek() {
    echo ""
    echo "=========================================="
    echo "[ENV 1] uipress-deepseek"
    echo "  transformers==4.46.3 (required by DeepSeek-OCR)"
    echo "=========================================="

    detect_cuda

    conda create -n uipress-deepseek python=3.11 -y 2>/dev/null || true

    # 使用 conda run 避免 conda activate 在脚本中的问题
    conda run -n uipress-deepseek pip install \
        torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 \
        --index-url "$TORCH_INDEX"

    conda run -n uipress-deepseek pip install \
        -r "$PROJ_DIR/requirements_deepseek.txt"

    # Flash Attention (optional but recommended)
    conda run -n uipress-deepseek pip install \
        flash-attn==2.7.3 --no-build-isolation 2>/dev/null || \
        echo "  [WARN] flash-attn install failed. Will use eager attention."

    echo "[OK] uipress-deepseek ready!"
}

setup_qwen() {
    echo ""
    echo "=========================================="
    echo "[ENV 2] uipress-qwen"
    echo "  transformers>=4.51 (for Qwen2.5-VL)"
    echo "=========================================="

    detect_cuda

    conda create -n uipress-qwen python=3.11 -y 2>/dev/null || true

    conda run -n uipress-qwen pip install \
        torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 \
        --index-url "$TORCH_INDEX"

    conda run -n uipress-qwen pip install \
        -r "$PROJ_DIR/requirements_qwen.txt"

    echo "[OK] uipress-qwen ready!"
    echo ""
    echo "  [Optional] For Qwen3-VL support, also run:"
    echo "    conda activate uipress-qwen"
    echo "    pip install git+https://github.com/huggingface/transformers"
}

case $TARGET in
    deepseek) setup_deepseek ;;
    qwen)     setup_qwen ;;
    all)      setup_deepseek; setup_qwen ;;
    *)        echo "Usage: $0 [deepseek|qwen|all]"; exit 1 ;;
esac

echo ""
echo "=========================================="
echo "Environment setup complete!"
echo ""
echo "Next: download data (either env works):"
echo "  conda activate uipress-qwen"
echo "  python scripts/download_data.py"
echo "=========================================="