first cimmit

.dockerignore

@@ -0,0 +1,2 @@
+models
+test

.gitignore

@@ -150,4 +150,5 @@ app/logs/*
 *.avi
 *.json
 *.env*
-config.backup.py
+config.backup.py
+models

Dockerfile

@@ -16,20 +16,21 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
         libgl1 \
         libglib2.0-0 \
         ca-certificates \
+        git \
     && apt-get clean \
     && rm -rf /var/lib/apt/lists/*
 
 WORKDIR /app
 
-COPY pyproject.toml uv.lock ./
+COPY pyproject.toml uv.lock download_model.sh ./
 
 ENV UV_COMPILE_BYTECODE=0
 
-RUN uv sync --frozen --no-dev --no-install-project --python 3.10
+RUN uv sync --frozen --no-dev --no-install-project --python 3.13
 
 # 4. 拷贝项目文件并安装项目本身
-COPY . .
+#COPY . .
-RUN uv sync --frozen --no-dev --python 3.10
+RUN uv sync --frozen --no-dev --python 3.13
 
 ENV PATH="/app/.venv/bin:$PATH"
 

app/config/config.py

@@ -17,7 +17,10 @@ class Settings(BaseSettings):
         extra='ignore'  # 忽略环境变量中多余的键
     )
     # 启动端口
-    SERVE_PROD: int = Field(default=8000, description='')
+    SERVE_PORT: int = Field(default=10090, description='')
+
+    FLUX2_KLEIN_MODEL_PATH: str = Field(default="/models", description='')
+
     # 调试配饰
     LOCAL: int = Field(default=0, description="是否在本地运行，1表示本地运行，0表示生产环境运行")
 

app/logging_env.py

@@ -1,4 +1,4 @@
-LOGS_PATH = 'logs/'
+LOGS_PATH = './logs/'
 LOGGER_CONFIG_DICT = {
     'version': 1,
     'disable_existing_loggers': False,

app/main.py

@@ -3,8 +3,7 @@ import os
 
 import litserve as ls
 
-from app.config.config import settings
+from app.logging_env import LOGGER_CONFIG_DICT
-from logging_env import LOGGER_CONFIG_DICT
 from app.server.generate_image.flux2_klein.server import Flux2KleinServer
 
 logger = logging.getLogger(__name__)
@@ -22,4 +21,4 @@ if __name__ == '__main__':
     to_product = Flux2KleinServer(api_path='/api/v1/to_product')
 
     server = ls.LitServer([to_product])
-    server.run(port=settings.SERVE_PROD)
+    server.run(port=8888)

app/server/generate_image/flux2_klein/client.py

@@ -9,5 +9,5 @@ request_data = {
     "infer_step":4,
     "tasks_id":"123456-123"
 }
-response = requests.post("http://127.0.0.1:8012//api/v1/to_product", json=request_data)
+response = requests.post("http://127.0.0.1:10090//api/v1/to_product", json=request_data)
 print(f"Status: {response.status_code}\nResponse:\n {response.text}")

app/server/generate_image/flux2_klein/server.py

@@ -1,4 +1,6 @@
 import io
+import os
+
 import torch
 import litserve as ls
 from diffusers import Flux2KleinPipeline
@@ -15,7 +17,8 @@ class Flux2KleinServer(ls.LitAPI):
         # Load the model
         dtype = torch.bfloat16
         self.device = device
-        self.model = Flux2KleinPipeline.from_pretrained("black-forest-labs/FLUX.2-klein-4B", torch_dtype=dtype, is_distilled=False)
+        model_path = os.path.join(settings.FLUX2_KLEIN_MODEL_PATH, "FLUX.2-klein-4B")
+        self.model = Flux2KleinPipeline.from_pretrained(model_path, torch_dtype=dtype, is_distilled=False)
         self.model.to(device)  # save some VRAM by offloading the model to CPU
 
     def decode_request(self, request):

client.py

@@ -1,7 +0,0 @@
-# This file is auto-generated by LitServe.
-# Disable auto-generation by setting `generate_client_file=False` in `LitServer.run()`.
-
-import requests
-
-response = requests.post("http://127.0.0.1:8012/predict", json={"input": 4.0})
-print(f"Status: {response.status_code}\nResponse:\n {response.text}")

docker-compose.yml

@@ -1,20 +1,17 @@
 services:
-  AiDA_Model_Litserve:
+  aida_model_litserve:
     build:
       context: .
       dockerfile: Dockerfile
     working_dir: /app
-    environment:
-      GOOGLE_APPLICATION_CREDENTIALS: /google_application_credentials.json
-      DEBUG: 0
     volumes:
       - ./app:/app/app
-      - ./.prod_env:/app/.env
+      - ./.env_prod:/app/.env
-      - ./data:/data
+      - ./models:/models
-      - ./google_application_credentials.json:/google_application_credentials.json
+      - ./logs:/app/logs
       - /etc/localtime:/etc/localtime:ro
     ports:
-      - "10070:8000"
+      - "${SERVE_PORT}:8888"
     deploy:
       resources:
         reservations:

download_model.sh

@@ -0,0 +1 @@
+huggingface-cli download black-forest-labs/FLUX.2-klein-4B --local-dir /data/models/FLUX.2-klein-4B