LiquidAI LFM3-VL image understanding with the Python SDK
LiquidAI LFM3-VL image understanding with the Python SDK.
Run with python examples/python/lfm3_vl.py photo.jpg.
The architecture is detected automatically; Q4K quantization reduces weight memory.
"""LiquidAI LFM3-VL image understanding with the Python SDK.
Run with `python examples/python/lfm3_vl.py photo.jpg`.The architecture is detected automatically; Q4K quantization reduces weight memory."""
import argparsefrom pathlib import Path
from mistralrs import ChatCompletionRequest, Runner, Which
parser = argparse.ArgumentParser(description=__doc__)parser.add_argument("image", type=Path, help="Path to an image")args = parser.parse_args()
runner = Runner( which=Which.MultimodalPlain(model_id="LiquidAI/LFM3-VL-2B"), in_situ_quant="q4k",)
res = runner.send_chat_completion_request( ChatCompletionRequest( model="default", messages=[ { "role": "user", "content": [ { "type": "image_url", "image_url": {"url": args.image.resolve().as_uri()}, }, { "type": "text", "text": "Describe this image and identify the main subject.", }, ], } ], max_tokens=256, ))print(res.choices[0].message.content)print(res.usage)Source: examples/python/lfm3_vl.py