colomi
42622e8e1e
- setup.sh: 自动检测 CONDA_BASE（conda info --base），移除 set -e 改为 错误收集，新增第8步验证 kohya 训练环境（测试 torch/accelerate 加载） - start_all_services.sh: 自动检测 CONDA_BASE，新增端口冲突检测和 WebUI 就绪等待（最多60秒轮询） - stop_all_services.sh: 新增停止脚本，按端口停止三个服务 - py310.yml: 移除硬编码 prefix 路径，避免跨机器部署问题 - DEPLOY.md: 详细部署操作手册，覆盖从 clone 到验证三大功能的全流程 - 已验证: kohya.tar.gz 解压→conda-unpack→PYTHONPATH 加载全部通过
131 lines
4.8 KiB
Bash
Executable File
131 lines
4.8 KiB
Bash
Executable File
#!/bin/bash
|
||
|
||
BASE_DIR="$(cd "$(dirname "$0")" && pwd)"
|
||
|
||
# 自动检测 conda 安装路径（优先用环境变量，其次用 conda info --base）
|
||
if [ -z "$CONDA_BASE" ]; then
|
||
if command -v conda >/dev/null 2>&1; then
|
||
CONDA_BASE="$(conda info --base 2>/dev/null)"
|
||
fi
|
||
fi
|
||
CONDA_BASE="${CONDA_BASE:-$HOME/miniconda3}"
|
||
|
||
# 初始化 conda（脚本中需要 source 才能用 conda activate）
|
||
if [ -f "$CONDA_BASE/etc/profile.d/conda.sh" ]; then
|
||
source "$CONDA_BASE/etc/profile.d/conda.sh"
|
||
else
|
||
echo "WARNING: 未找到 conda.sh，请确认 CONDA_BASE=$CONDA_BASE 正确"
|
||
fi
|
||
|
||
echo "=== 换发型项目部署脚本 ==="
|
||
echo "BASE_DIR: $BASE_DIR"
|
||
echo "CONDA_BASE: $CONDA_BASE"
|
||
echo ""
|
||
|
||
ERRORS=()
|
||
|
||
# 1. 检查前提条件
|
||
echo "[1/8] 检查前提条件..."
|
||
command -v git >/dev/null || { echo " ✗ git 未安装"; ERRORS+=("git"); }
|
||
command -v conda >/dev/null || { echo " ✗ conda 未安装"; ERRORS+=("conda"); }
|
||
nvidia-smi >/dev/null 2>&1 || { echo " ✗ NVIDIA 驱动未安装"; ERRORS+=("nvidia-smi"); }
|
||
if [ ${#ERRORS[@]} -eq 0 ]; then
|
||
echo " ✓ 前提条件满足"
|
||
fi
|
||
|
||
# 2. 生成 configure.ini
|
||
echo "[2/8] 生成 configure.ini..."
|
||
sed "s|__BASE_DIR__|$BASE_DIR|g" "$BASE_DIR/hair_service_sd/config/configure.ini.template" > "$BASE_DIR/hair_service_sd/config/configure.ini"
|
||
echo " ✓ configure.ini 已生成"
|
||
|
||
# 3. 恢复 conda 环境（conda-pack）
|
||
echo "[3/8] 恢复 conda 环境（my_hair, sdwebui, kohya）..."
|
||
for env_name in my_hair sdwebui kohya; do
|
||
if [ -f "$BASE_DIR/conda_envs/${env_name}.tar.gz" ]; then
|
||
echo " 恢复 $env_name ..."
|
||
rm -rf "$CONDA_BASE/envs/$env_name"
|
||
mkdir -p "$CONDA_BASE/envs/$env_name"
|
||
tar -xzf "$BASE_DIR/conda_envs/${env_name}.tar.gz" -C "$CONDA_BASE/envs/$env_name"
|
||
conda activate "$env_name" 2>/dev/null
|
||
conda-unpack 2>/dev/null # 修复打包后的硬编码路径
|
||
conda deactivate 2>/dev/null
|
||
echo " ✓ $env_name 已恢复"
|
||
else
|
||
echo " ✗ conda_envs/${env_name}.tar.gz 不存在，请从网盘下载"
|
||
ERRORS+=("conda_envs/${env_name}.tar.gz")
|
||
fi
|
||
done
|
||
|
||
# 4. 创建 py310 环境（从 yml）
|
||
echo "[4/8] 创建 py310 环境（从 yml）..."
|
||
if [ -f "$BASE_DIR/conda_envs/py310.yml" ]; then
|
||
conda env create -f "$BASE_DIR/conda_envs/py310.yml" -n py310 2>/dev/null && echo " ✓ py310 已创建" || echo " ℹ py310 环境已存在，跳过"
|
||
else
|
||
echo " ✗ conda_envs/py310.yml 不存在"
|
||
ERRORS+=("conda_envs/py310.yml")
|
||
fi
|
||
|
||
# 5. 检查模型/数据目录
|
||
echo "[5/8] 检查模型和数据目录..."
|
||
for dir in hair_service_sd/weights stable-diffusion-webui/models/Lora stable-diffusion-webui/models/Stable-diffusion stable-diffusion-webui/extensions/sd-webui-controlnet stable-diffusion-webui/repositories kohya_ss_home/.local kohya_ss_home/kohya_ss data/ref_hairstyle; do
|
||
if [ -d "$BASE_DIR/$dir" ]; then
|
||
echo " ✓ $dir"
|
||
else
|
||
echo " ✗ $dir 缺失，请从网盘下载"
|
||
ERRORS+=("$dir")
|
||
fi
|
||
done
|
||
|
||
# 6. 检查训练底模
|
||
echo "[6/8] 检查训练底模 v1-5-pruned-emaonly..."
|
||
if [ -f "$BASE_DIR/stable-diffusion-webui/models/Stable-diffusion/v1-5-pruned-emaonly.safetensors" ]; then
|
||
echo " ✓ v1-5-pruned-emaonly.safetensors 已存在"
|
||
else
|
||
echo " ✗ v1-5-pruned-emaonly.safetensors 缺失，请从网盘下载 models/ 目录"
|
||
ERRORS+=("v1-5-pruned-emaonly.safetensors")
|
||
fi
|
||
|
||
# 7. 创建运行时目录
|
||
echo "[7/8] 创建运行时目录..."
|
||
mkdir -p "$BASE_DIR/logs"
|
||
mkdir -p "$BASE_DIR/data/tmp" "$BASE_DIR/data/res_dir" "$BASE_DIR/data/userImage" "$BASE_DIR/data/user_info"
|
||
mkdir -p "$BASE_DIR/kohya_ss_home/train_material"
|
||
echo " ✓ 运行时目录已创建"
|
||
|
||
# 8. 验证 kohya 训练环境
|
||
echo "[8/8] 验证 kohya 训练环境..."
|
||
KOHYA_PYTHON="$CONDA_BASE/envs/kohya/bin/python"
|
||
KOHYA_LOCAL="$BASE_DIR/kohya_ss_home/.local/lib/python3.10/site-packages"
|
||
if [ -f "$KOHYA_PYTHON" ] && [ -d "$KOHYA_LOCAL" ]; then
|
||
echo " 测试 torch/accelerate 加载..."
|
||
if PYTHONPATH="$KOHYA_LOCAL" "$KOHYA_PYTHON" -c "
|
||
import torch, accelerate, bitsandbytes, diffusers
|
||
print(f' torch={torch.__version__} cuda={torch.cuda.is_available()}')
|
||
print(f' accelerate={accelerate.__version__}')
|
||
print(' ✓ 训练环境验证通过')
|
||
" 2>/dev/null; then
|
||
true
|
||
else
|
||
echo " ✗ 训练环境验证失败，请检查 kohya 环境和 .local 目录"
|
||
ERRORS+=("kohya验证")
|
||
fi
|
||
else
|
||
echo " ✗ kohya Python 或 .local 目录不存在，跳过验证"
|
||
ERRORS+=("kohya环境")
|
||
fi
|
||
|
||
echo ""
|
||
echo "=== 部署结果 ==="
|
||
if [ ${#ERRORS[@]} -eq 0 ]; then
|
||
echo "✓ 所有检查通过，部署完成！"
|
||
else
|
||
echo "⚠ 有 ${#ERRORS[@]} 个问题需要处理："
|
||
for err in "${ERRORS[@]}"; do
|
||
echo " - $err"
|
||
done
|
||
echo ""
|
||
echo "请从网盘下载缺失的文件后重新运行 ./setup.sh"
|
||
fi
|
||
echo ""
|
||
echo "启动服务: ./start_all_services.sh"
|