app.py

import torch
from PIL import Image
import requests
from io import BytesIO  # Importing BytesIO from the io module
from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info

# Check if CUDA is available and set the device accordingly
device = "cuda" if torch.cuda.is_available() else "cpu"

# Load the model on the available device
model = Qwen2VLForConditionalGeneration.from_pretrained(
    "Qwen/Qwen2-VL-2B-Instruct", torch_dtype="auto", device_map="auto"
)
model = model.to(device)

# Default processor
processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")

# Resize the image to a smaller resolution (e.g., 512x512)
image_url = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
response = requests.get(image_url)
img = Image.open(BytesIO(response.content))  # Using BytesIO to handle image from the byte stream

# Resize the image
img_resized = img.resize((512, 512))  # Resize the image to 512x512
image_inputs = processor(images=img_resized, return_tensors="pt").to(device)

# Prepare the text input
messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": img_resized,
            },
            {"type": "text", "text": "Describe this image."},
        ],
    }
]

# Preparation for inference
text = processor.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True
)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
    text=[text],
    images=image_inputs,
    videos=video_inputs,
    padding=True,
    return_tensors="pt",
)
inputs = inputs.to(device)  # Move inputs to the same device as the model

# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)

# Trim the output tokens
generated_ids_trimmed = [
    out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]

# Decode the generated text
output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)

# Print the output
print(output_text)