Add smoke test, initial serving tasks

Signed-off-by: jphillips <[email protected]>

modules/odr_caption/.env.caption.template

@@ -0,0 +1,5 @@
+VLLM_HOST=0.0.0.0
+VLLM_PORT=32000
+
+CAPTION_API_HOST=0.0.0.0
+CAPTION_API_PORT=32001

modules/odr_caption/Taskfile.yml

@@ -0,0 +1,14 @@
+version: '3'
+
+tasks:
+  qwen-7b:
+    cmds:
+      - python -m vllm.entrypoints.openai.api_server --model Qwen/Qwen2-VL-7B-Instruct  --host 0.0.0.0 --port 32000
+
+  transformers:
+    cmds:
+      - pip install git+https://github.com/huggingface/transformers.git --upgrade
+
+  qwen:
+    cmds:
+      -  vllm serve Qwen/Qwen2-VL-7B-Instruct-GPTQ-Int8

modules/odr_caption/requirements.txt

@@ -0,0 +1,12 @@
+vllm
+transformers
+qwen-vl-utils[decord]
+fastapi
+click
+openai
+pydantic
+termcolor
+pillow
+flash-attn
+python-dotenv
+vllm>=0.6.2

modules/odr_caption/test/NOTICE.md

@@ -0,0 +1 @@
+All images in this folder should be available under open licenses.

modules/odr_caption/test/smoke.py

@@ -0,0 +1,39 @@
+import os
+import asyncio
+from dotenv import load_dotenv
+from odr_caption.agents.ImageCaptioner import ImageCaptioner
+from odr_caption.utils.logger import logger
+
+load_dotenv()
+
+# Set vLLM's API server URL
+vllm_api_base = "http://localhost:32000/v1"
+
+
+async def main():
+    cwd = os.getcwd()
+    image_path = f"{cwd}/test/test_images/test_image_1.png"
+
+    # Initialize ImageCaptioner
+    captioner = ImageCaptioner(
+        vllm_server_url=vllm_api_base,
+        model_name="Qwen/Qwen2-7B-Instruct",
+        max_tokens=2048,
+        temperature=0.35,
+    )
+
+    # Define system message and prompt
+    system_message = "You are a helpful assistant."
+    prompt = "What is the text in the illustration?"
+
+    # Caption the image
+    response = await captioner.caption_image(image_path, system_message, prompt)
+
+    # Print the response
+    logger.info(f"Chat response: {response}")
+    if response.choices:
+        logger.info(f"Generated caption: {response.choices[0].message.content}")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())