Qwen2-vl:No chat template is set for this processor. Please either set the `chat_template` attribute

peft库微调大模型后合并推理问题排查

先将lora合并到base模型中

from transformers import AutoTokenizer, Qwen2VLForConditionalGeneration
from peft import PeftModel
import torch

base_model_path = "E:\ModelScopeModels\hub\Qwen\Qwen2-VL-2B-Instruct"
lora_model_path = "E:\LLm_finetuning\Qwen2-VL-2B-LatexOCR\checkpoint-124"

merged_model_path = "E:\ModelScopeModels\hub\Qwen\Qwen2-VL-2B-Instruct-LaTexOCR-merge"
#加载模型和分词器
model = Qwen2VLForConditionalGeneration.from_pretrained(base_model_path,torch_dtype=torch.float16,trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(base_model_path)
print(model)

#加载lora适配器
model = PeftModel.from_pretrained(model,lora_model_path)
#合并权重
merged_model = model.merge_and_unload()

#保存合并后的模型
merged_model.save_pretrained(merged_model_path)
tokenizer.save_pretrained(merged_model_path)
print(f"merged model saved to {merged_model_path}")

inference

from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info
from peft import PeftModel, LoraConfig, TaskType
prompt = "你是一个LaText OCR助手,目标是读取用户输入的照片，转换成LaTex公式。"
merged_model_path = r"xxxxxx\Qwen\Qwen2-VL-2B-Instruct-merge"
test_image_path = "997.jpg"

config = LoraConfig(
    task_type=TaskType.CAUSAL_LM,
    target_modules=["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
    inference_mode=True,
    r=64,  # Lora 秩
    lora_alpha=16,  # Lora alaph，具体作用参见 Lora 原理
    lora_dropout=0.05,  # Dropout 比例
    bias="none",
)

# default: Load the model on the available device(s)
model = Qwen2VLForConditionalGeneration.from_pretrained(
    merged_model_path, torch_dtype="auto", device_map="auto"
)
processor = AutoProcessor.from_pretrained(merged_model_path)
messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": test_image_path,
                "resized_height": 100,
                "resized_width": 500,
            },
            {"type": "text", "text": f"{prompt}"},
        ],
    }
]
# 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=8192)
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[0])

合并后，用合并后的模型进行推理，可能会出现以下错误：No chat template is set for this processor.

原因是合并后的模型文件夹内缺失chat_template.json文件，需要从原基础模型中复制一份到合并后的文件夹内。