TBAC_test.py

from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
import argparse

# 设置命令行参数
parser = argparse.ArgumentParser(description='TBAC测试脚本')
parser.add_argument('image_path', nargs='?', type=str, help='图像文件路径(第一位置参数)')
parser.add_argument('-i', '--image_path', dest='image_path', type=str, help='图像文件路径')
parser.add_argument('query', nargs='?', type=str, default='告诉我图中有什么？', help='查询问题(第二位置参数)')
parser.add_argument('-q', '--query', dest='query', type=str, default='告诉我图中有什么？', help='查询问题')
args = parser.parse_args()

# 验证必填参数
if args.image_path is None:
    parser.error("必须提供图像路径，可通过位置参数或-i/--image_path选项指定")

model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "./TBAC-VLR1-3B-preview", torch_dtype="auto"
)

processor = AutoProcessor.from_pretrained("./TBAC-VLR1-3B-preview")

# 使用命令行参数
image_path = args.image_path
query = args.query

messages = [
    {
        "role": "system",
        "content": "You are a helpful assistant. The user asks a question, and you solve it. You need first think about the reasoning process in the mind and then provides the user with the answer. The answer are enclosed within \\boxed{} tags i.e., reasoning process here \\boxed{ answer here }."
    },
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": image_path,
            },
            {"type": "text", "text": query},
        ],
    }
]

# 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("cuda")

# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=1024*4)
generated_ids_trimmed = [
    out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
output_text = processor.batch_decode(
    generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print(output_text)