Dedicated docker-compose.yml for gpu deployment

.dockerignore

@@ -0,0 +1,75 @@
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+*.egg-info/
+.eggs/
+dist/
+build/
+*.egg
+
+# Virtual environments
+.venv/
+venv/
+ENV/
+env/
+.pixi/
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+.cursor/
+
+# Git
+.git/
+.gitignore
+
+# Data (不需要打包到镜像)
+data/
+!data/.gitkeep
+
+# Notebooks
+notebooks/
+*.ipynb
+
+# Documentation
+docs/
+
+# Reports
+reports/
+
+# References
+references/
+
+# Scripts (训练脚本不需要)
+scripts/
+
+# Lock files
+pixi.lock
+
+# Tests
+tests/
+.pytest_cache/
+
+# Logs
+*.log
+logs/
+
+# Temporary files
+*.tmp
+*.temp
+.DS_Store
+
+# Models (will be mounted as volume or copied explicitly)
+# Note: models/final/ is copied in Dockerfile
+models/finetune_cv/
+models/pretrain_cv/
+models/mpnn/
+models/*.pt
+models/*.json
+!models/final/
+

Dockerfile

@@ -0,0 +1,63 @@
+# LNP-ML Docker Image
+# 多阶段构建，支持 API 和 Streamlit 两种服务
+
+FROM python:3.8-slim AS base
+
+# 设置环境变量
+ENV PYTHONUNBUFFERED=1 \
+    PYTHONDONTWRITEBYTECODE=1 \
+    PIP_NO_CACHE_DIR=1 \
+    PIP_DISABLE_PIP_VERSION_CHECK=1
+
+WORKDIR /app
+
+# 安装系统依赖
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential \
+    libxrender1 \
+    libxext6 \
+    curl \
+    && rm -rf /var/lib/apt/lists/*
+
+# 复制依赖文件
+COPY requirements.txt .
+
+# 安装 Python 依赖
+RUN pip install --upgrade pip && \
+    pip install -r requirements.txt
+
+# 复制项目代码
+COPY pyproject.toml .
+COPY README.md .
+COPY LICENSE .
+COPY lnp_ml/ ./lnp_ml/
+COPY app/ ./app/
+
+# 安装项目包
+RUN pip install -e .
+
+# 复制模型文件
+COPY models/final/ ./models/final/
+
+# ============ API 服务 ============
+FROM base AS api
+
+EXPOSE 8000
+
+ENV MODEL_PATH=/app/models/final/model.pt
+
+CMD ["uvicorn", "app.api:app", "--host", "0.0.0.0", "--port", "8000"]
+
+# ============ Streamlit 服务 ============
+FROM base AS streamlit
+
+EXPOSE 8501
+
+# Streamlit 配置
+ENV STREAMLIT_SERVER_PORT=8501 \
+    STREAMLIT_SERVER_ADDRESS=0.0.0.0 \
+    STREAMLIT_SERVER_HEADLESS=true \
+    STREAMLIT_BROWSER_GATHER_USAGE_STATS=false
+
+CMD ["streamlit", "run", "app/app.py"]
+

Makefile

@@ -200,6 +200,48 @@ serve:
 	@echo "然后访问: http://localhost:8501"
 
 
+#################################################################################
+# DOCKER COMMANDS                                                               #
+#################################################################################
+
+## Build Docker images
+.PHONY: docker-build
+docker-build:
+	docker compose build
+
+## Start all services with Docker Compose
+.PHONY: docker-up
+docker-up:
+	docker compose up -d
+
+## Stop all Docker services
+.PHONY: docker-down
+docker-down:
+	docker compose down
+
+## View Docker logs
+.PHONY: docker-logs
+docker-logs:
+	docker compose logs -f
+
+## Build and start all services
+.PHONY: docker-serve
+docker-serve: docker-build docker-up
+	@echo ""
+	@echo "🚀 服务已启动!"
+	@echo "   - API:       http://localhost:8000"
+	@echo "   - Web 应用:  http://localhost:8501"
+	@echo ""
+	@echo "查看日志: make docker-logs"
+	@echo "停止服务: make docker-down"
+
+## Clean Docker resources (images, volumes, etc.)
+.PHONY: docker-clean
+docker-clean:
+	docker compose down -v --rmi local
+	docker system prune -f
+
+
 #################################################################################
 # Self Documenting Commands                                                     #
 #################################################################################

app/app.py

@@ -3,9 +3,13 @@ Streamlit 配方优化交互界面
 
 启动应用:
     streamlit run app/app.py
+
+Docker 环境变量:
+    API_URL: API 服务地址 (默认: http://localhost:8000)
 """
 
 import io
+import os
 from datetime import datetime
 
 import httpx
@@ -14,7 +18,8 @@ import streamlit as st
 
 # ============ 配置 ============
 
-API_URL = "http://localhost:8000"
+# 从环境变量读取 API 地址，支持 Docker 环境
+API_URL = os.environ.get("API_URL", "http://localhost:8000")
 
 AVAILABLE_ORGANS = [
     "liver",
@@ -27,13 +32,13 @@ AVAILABLE_ORGANS = [
 ]
 
 ORGAN_LABELS = {
-    "liver": "🫀 肝脏 (Liver)",
-    "spleen": "🟣 脾脏 (Spleen)",
-    "lung": "🫁 肺 (Lung)",
-    "heart": "❤️ 心脏 (Heart)",
-    "kidney": "🫘 肾脏 (Kidney)",
-    "muscle": "💪 肌肉 (Muscle)",
-    "lymph_nodes": "🔵 淋巴结 (Lymph Nodes)",
+    "liver": "肝脏 (Liver)",
+    "spleen": "脾脏 (Spleen)",
+    "lung": "肺 (Lung)",
+    "heart": "心脏 (Heart)",
+    "kidney": "肾脏 (Kidney)",
+    "muscle": "肌肉 (Muscle)",
+    "lymph_nodes": "淋巴结 (Lymph Nodes)",
 }
 
 # ============ 页面配置 ============
@@ -146,19 +151,20 @@ def format_results_dataframe(results: dict) -> pd.DataFrame:
     for f in formulations:
         row = {
             "排名": f["rank"],
-            f"Biodist_{target_organ}": f"{f['target_biodist']:.4f}",
-            "阳离子脂质/mRNA": f["cationic_lipid_to_mrna_ratio"],
-            "阳离子脂质(mol)": f["cationic_lipid_mol_ratio"],
-            "磷脂(mol)": f["phospholipid_mol_ratio"],
-            "胆固醇(mol)": f["cholesterol_mol_ratio"],
-            "PEG脂质(mol)": f["peg_lipid_mol_ratio"],
+            # f"{target_organ}分布": f"{f['target_biodist']*100:.2f}%",
+            f"{target_organ}分布": f"{f['target_biodist']*100:.8f}%",
+            "阳离子脂质/mRNA比例": f["cationic_lipid_to_mrna_ratio"],
+            "阳离子脂质(mol)比例": f["cationic_lipid_mol_ratio"],
+            "磷脂(mol)比例": f["phospholipid_mol_ratio"],
+            "胆固醇(mol)比例": f["cholesterol_mol_ratio"],
+            "PEG脂质(mol)比例": f["peg_lipid_mol_ratio"],
             "辅助脂质": f["helper_lipid"],
             "给药途径": f["route"],
         }
         # 添加其他器官的 biodist
         for organ, value in f["all_biodist"].items():
             if organ != target_organ:
-                row[f"Biodist_{organ}"] = f"{value:.4f}"
+                row[f"{organ}分布"] = f"{value*100:.2f}%"
         rows.append(row)
     
     return pd.DataFrame(rows)
@@ -184,7 +190,7 @@ def main():
     
     # ========== 侧边栏 ==========
     with st.sidebar:
-        st.header("⚙️ 参数设置")
+        # st.header("⚙️ 参数设置")
         
         # API 状态
         if api_online:
@@ -193,7 +199,7 @@ def main():
             st.error("🔴 API 服务离线")
             st.info("请先启动 API 服务:\n```\nuvicorn app.api:app --port 8000\n```")
         
-        st.divider()
+        # st.divider()
         
         # SMILES 输入
         st.subheader("🔬 分子结构")
@@ -206,18 +212,18 @@ def main():
         )
         
         # 示例 SMILES
-        with st.expander("📋 示例 SMILES"):
-            example_smiles = {
-                "DLin-MC3-DMA": "CC(C)=CCCC(C)=CCCC(C)=CCN(C)CCCCCCCCOC(=O)CCCCCCC/C=C\\CCCCCCCC",
-                "简单胺": "CC(C)NCCNC(C)C",
-                "长链胺": "CCCCCCCCCCCCNCCNCCCCCCCCCCCC",
-            }
-            for name, smi in example_smiles.items():
-                if st.button(f"使用 {name}", key=f"example_{name}"):
-                    st.session_state["smiles_input"] = smi
-                    st.rerun()
+        # with st.expander("📋 示例 SMILES"):
+        #     example_smiles = {
+        #         "DLin-MC3-DMA": "CC(C)=CCCC(C)=CCCC(C)=CCN(C)CCCCCCCCOC(=O)CCCCCCC/C=C\\CCCCCCCC",
+        #         "简单胺": "CC(C)NCCNC(C)C",
+        #         "长链胺": "CCCCCCCCCCCCNCCNCCCCCCCCCCCC",
+        #     }
+        #     for name, smi in example_smiles.items():
+        #         if st.button(f"使用 {name}", key=f"example_{name}"):
+        #             st.session_state["smiles_input"] = smi
+        #             st.rerun()
         
-        st.divider()
+        # st.divider()
         
         # 目标器官选择
         st.subheader("🎯 目标器官")
@@ -228,7 +234,7 @@ def main():
             index=0,
         )
         
-        st.divider()
+        # st.divider()
         
         # 高级选项
         with st.expander("🔧 高级选项"):
@@ -294,8 +300,8 @@ def main():
         with col2:
             best_score = results["formulations"][0]["target_biodist"]
             st.metric(
-                "最优 Biodistribution",
-                f"{best_score:.4f}",
+                "最优分布",
+                f"{best_score*100:.2f}%",
             )
         
         with col3:
@@ -333,61 +339,61 @@ def main():
         )
         
         # 详细信息
-        with st.expander("🔍 查看最优配方详情"):
-            best = results["formulations"][0]
+        # with st.expander("🔍 查看最优配方详情"):
+        #     best = results["formulations"][0]
             
-            col1, col2 = st.columns(2)
+        #     col1, col2 = st.columns(2)
             
-            with col1:
-                st.markdown("**配方参数**")
-                st.json({
-                    "阳离子脂质/mRNA 比例": best["cationic_lipid_to_mrna_ratio"],
-                    "阳离子脂质 (mol%)": best["cationic_lipid_mol_ratio"],
-                    "磷脂 (mol%)": best["phospholipid_mol_ratio"],
-                    "胆固醇 (mol%)": best["cholesterol_mol_ratio"],
-                    "PEG 脂质 (mol%)": best["peg_lipid_mol_ratio"],
-                    "辅助脂质": best["helper_lipid"],
-                    "给药途径": best["route"],
-                })
+        #     with col1:
+        #         st.markdown("**配方参数**")
+        #         st.json({
+        #             "阳离子脂质/mRNA 比例": best["cationic_lipid_to_mrna_ratio"],
+        #             "阳离子脂质 (mol%)": best["cationic_lipid_mol_ratio"],
+        #             "磷脂 (mol%)": best["phospholipid_mol_ratio"],
+        #             "胆固醇 (mol%)": best["cholesterol_mol_ratio"],
+        #             "PEG 脂质 (mol%)": best["peg_lipid_mol_ratio"],
+        #             "辅助脂质": best["helper_lipid"],
+        #             "给药途径": best["route"],
+        #         })
             
-            with col2:
-                st.markdown("**各器官 Biodistribution 预测**")
-                biodist_df = pd.DataFrame([
-                    {"器官": ORGAN_LABELS.get(k, k), "Biodistribution": f"{v:.4f}"}
-                    for k, v in best["all_biodist"].items()
-                ])
-                st.dataframe(biodist_df, hide_index=True, use_container_width=True)
+        #     with col2:
+        #         st.markdown("**各器官 Biodistribution 预测**")
+        #         biodist_df = pd.DataFrame([
+        #             {"器官": ORGAN_LABELS.get(k, k), "Biodistribution": f"{v:.4f}"}
+        #             for k, v in best["all_biodist"].items()
+        #         ])
+        #         st.dataframe(biodist_df, hide_index=True, use_container_width=True)
     
     else:
         # 欢迎信息
         st.info("👈 请在左侧输入 SMILES 并选择目标器官，然后点击「开始配方优选」")
         
         # 使用说明
-        with st.expander("📖 使用说明"):
-            st.markdown("""
-            ### 如何使用
+        # with st.expander("📖 使用说明"):
+        #     st.markdown("""
+        #     ### 如何使用
             
-            1. **输入 SMILES**: 在左侧输入框中输入阳离子脂质的 SMILES 字符串
-            2. **选择目标器官**: 选择您希望优化的器官靶向
-            3. **点击优选**: 系统将自动搜索最优配方组合
-            4. **查看结果**: 右侧将显示 Top-20 优选配方
-            5. **导出数据**: 点击导出按钮将结果保存为 CSV 文件
+        #     1. **输入 SMILES**: 在左侧输入框中输入阳离子脂质的 SMILES 字符串
+        #     2. **选择目标器官**: 选择您希望优化的器官靶向
+        #     3. **点击优选**: 系统将自动搜索最优配方组合
+        #     4. **查看结果**: 右侧将显示 Top-20 优选配方
+        #     5. **导出数据**: 点击导出按钮将结果保存为 CSV 文件
             
-            ### 优化参数
+        #     ### 优化参数
             
-            系统会优化以下配方参数:
-            - **阳离子脂质/mRNA 比例**: 0.05 - 0.30
-            - **阳离子脂质 mol 比例**: 0.05 - 0.80
-            - **磷脂 mol 比例**: 0.00 - 0.80
-            - **胆固醇 mol 比例**: 0.00 - 0.80
-            - **PEG 脂质 mol 比例**: 0.00 - 0.05
-            - **辅助脂质**: DOPE / DSPC / DOTAP
-            - **给药途径**: 静脉注射 / 肌肉注射
+        #     系统会优化以下配方参数:
+        #     - **阳离子脂质/mRNA 比例**: 0.05 - 0.30
+        #     - **阳离子脂质 mol 比例**: 0.05 - 0.80
+        #     - **磷脂 mol 比例**: 0.00 - 0.80
+        #     - **胆固醇 mol 比例**: 0.00 - 0.80
+        #     - **PEG 脂质 mol 比例**: 0.00 - 0.05
+        #     - **辅助脂质**: DOPE / DSPC / DOTAP
+        #     - **给药途径**: 静脉注射 / 肌肉注射
             
-            ### 约束条件
+        #     ### 约束条件
             
-            mol 比例之和 = 1 (阳离子脂质 + 磷脂 + 胆固醇 + PEG 脂质)
-            """)
+        #     mol 比例之和 = 1 (阳离子脂质 + 磷脂 + 胆固醇 + PEG 脂质)
+        #     """)
 
 
 if __name__ == "__main__":

docker-compose-gpu.yml

@@ -0,0 +1,54 @@
+services:
+  # FastAPI 后端服务
+  api:
+    build:
+      context: .
+      dockerfile: Dockerfile
+      target: api
+    container_name: lnp-api
+    environment:
+      - MODEL_PATH=/app/models/final/model.pt
+    volumes:
+      # 挂载模型目录以便更新模型
+      - ./models/final:/app/models/final:ro
+      - ./models/mpnn:/app/models/mpnn:ro
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:8000/"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 60s
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: 1
+              capabilities: [gpu]
+
+  # Streamlit 前端服务
+  streamlit:
+    build:
+      context: .
+      dockerfile: Dockerfile
+      target: streamlit
+    container_name: lnp-streamlit
+    ports:
+      - "8501:8501"
+    environment:
+      - API_URL=http://api:8000
+    depends_on:
+      api:
+        condition: service_started
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:8501/_stcore/health"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 30s
+
+networks:
+  default:
+    name: lnp-network

docker-compose.yml

@@ -0,0 +1,49 @@
+services:
+  # FastAPI 后端服务
+  api:
+    build:
+      context: .
+      dockerfile: Dockerfile
+      target: api
+    container_name: lnp-api
+    environment:
+      - MODEL_PATH=/app/models/final/model.pt
+      - DEVICE=cpu
+    volumes:
+      # 挂载模型目录以便更新模型
+      - ./models/final:/app/models/final:ro
+      - ./models/mpnn:/app/models/mpnn:ro
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:8000/"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 60s
+
+  # Streamlit 前端服务
+  streamlit:
+    build:
+      context: .
+      dockerfile: Dockerfile
+      target: streamlit
+    container_name: lnp-streamlit
+    ports:
+      - "8501:8501"
+    environment:
+      - API_URL=http://api:8000
+    depends_on:
+      api:
+        condition: service_started
+    restart: unless-stopped
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:8501/_stcore/health"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 30s
+
+networks:
+  default:
+    name: lnp-network
+

requirements.txt

@@ -0,0 +1,20 @@
+# 严格遵循 pixi.toml 的依赖版本
+# 注意: lnp_ml 本地包在 Dockerfile 中单独安装
+
+# conda dependencies (in pixi.toml [dependencies])
+loguru
+tqdm
+typer
+
+# pypi dependencies (in pixi.toml [pypi-dependencies])
+chemprop==1.7.0
+setuptools
+pandas>=2.0.3,<3
+openpyxl>=3.1.5,<4
+python-dotenv>=1.0.1,<2
+pyarrow>=17.0.0,<18
+fastparquet>=2024.2.0,<2025
+fastapi>=0.124.4,<0.125
+streamlit>=1.40.1,<2
+httpx>=0.28.1,<0.29
+uvicorn>=0.33.0,<0.34