完善部署脚本兼容性并生成详细部署手册

- 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

start_all_services.sh

@@ -1,24 +1,87 @@
 #!/bin/bash
+
 BASE_DIR="$(cd "$(dirname "$0")" && pwd)"
-CONDA_BASE="${CONDA_BASE:-/home/szlc/miniconda3}"
+
+# 自动检测 conda 安装路径
+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}"
 export CONDA_BASE
 
 mkdir -p "$BASE_DIR/logs"
 
+# 检查端口是否被占用，如占用则提示
+check_port() {
+    local port=$1
+    local name=$2
+    if lsof -i :$port >/dev/null 2>&1; then
+        echo "⚠ 端口 $port ($name) 已被占用，可能服务已在运行"
+        echo "  如需重启，请先运行: ./stop_all_services.sh"
+        return 1
+    fi
+    return 0
+}
+
+echo "=== 启动换发型服务 ==="
+echo "BASE_DIR: $BASE_DIR"
+echo "CONDA_BASE: $CONDA_BASE"
+echo ""
+
+# 检查 conda 环境是否存在
+for env_name in my_hair sdwebui py310; do
+    if [ ! -f "$CONDA_BASE/envs/$env_name/bin/python" ]; then
+        echo "✗ conda 环境 $env_name 不存在，请先运行 ./setup.sh"
+        exit 1
+    fi
+done
+
 # 1. hair_service_sd (端口 8801)
-cd "$BASE_DIR/hair_service_sd"
-nohup "$CONDA_BASE/envs/my_hair/bin/python" run_copy_cost_colorb64.py > "$BASE_DIR/logs/hair_service.log" 2>&1 &
-echo "hair_service_sd 已启动 (PID: $!, 端口 8801)"
+echo "[1/3] 启动 hair_service_sd (端口 8801)..."
+if check_port 8801 "hair_service_sd"; then
+    cd "$BASE_DIR/hair_service_sd"
+    nohup "$CONDA_BASE/envs/my_hair/bin/python" run_copy_cost_colorb64.py > "$BASE_DIR/logs/hair_service.log" 2>&1 &
+    echo "  hair_service_sd 已启动 (PID: $!, 端口 8801)"
+fi
 
 # 2. stable-diffusion-webui (端口 57860)
-cd "$BASE_DIR/stable-diffusion-webui"
-nohup "$CONDA_BASE/envs/sdwebui/bin/python" webui.py --api --listen --xformers --port 57860 > "$BASE_DIR/logs/webui.log" 2>&1 &
-echo "stable-diffusion-webui 已启动 (PID: $!, 端口 57860)"
+echo "[2/3] 启动 stable-diffusion-webui (端口 57860)..."
+if check_port 57860 "webui"; then
+    cd "$BASE_DIR/stable-diffusion-webui"
+    nohup "$CONDA_BASE/envs/sdwebui/bin/python" webui.py --api --listen --xformers --port 57860 > "$BASE_DIR/logs/webui.log" 2>&1 &
+    echo "  stable-diffusion-webui 已启动 (PID: $!, 端口 57860)"
+    echo "  ℹ WebUI 首次启动需加载模型（约 30-60 秒），请耐心等待"
+fi
 
 # 3. photo_service (端口 32678)
-cd "$BASE_DIR/photo_service"
-nohup "$CONDA_BASE/envs/py310/bin/python" lora_train_service_1.py > "$BASE_DIR/logs/photo_service.log" 2>&1 &
-echo "photo_service 已启动 (PID: $!, 端口 32678)"
+echo "[3/3] 启动 photo_service (端口 32678)..."
+if check_port 32678 "photo_service"; then
+    cd "$BASE_DIR/photo_service"
+    nohup "$CONDA_BASE/envs/py310/bin/python" lora_train_service_1.py > "$BASE_DIR/logs/photo_service.log" 2>&1 &
+    echo "  photo_service 已启动 (PID: $!, 端口 32678)"
+fi
 
 echo ""
-echo "所有服务已在后台启动，日志在 $BASE_DIR/logs/"
+echo "=== 所有服务已启动 ==="
+echo "日志目录: $BASE_DIR/logs/"
+echo ""
+echo "查看日志:"
+echo "  tail -f $BASE_DIR/logs/hair_service.log   # 换发型/换发色"
+echo "  tail -f $BASE_DIR/logs/webui.log          # SD WebUI"
+echo "  tail -f $BASE_DIR/logs/photo_service.log  # 训练服务"
+echo ""
+echo "停止服务: ./stop_all_services.sh"
+echo ""
+echo "等待 WebUI 就绪（检查端口 57860）..."
+for i in $(seq 1 60); do
+    if curl -s http://127.0.0.1:57860/sdapi/v1/options >/dev/null 2>&1; then
+        echo "✓ WebUI 已就绪"
+        break
+    fi
+    sleep 2
+    if [ $i -eq 60 ]; then
+        echo "⚠ WebUI 60 秒内未就绪，请检查日志: tail -f $BASE_DIR/logs/webui.log"
+    fi
+done