Delete image-info-train-count.py

image-info-train-count.py

@@ -1,62 +0,0 @@
-from transformers import (
-    AutoConfig,
-    AutoTokenizer,
-    BitsAndBytesConfig,
-    AutoProcessor, 
-    LlamaForCausalLM,
-    MllamaForConditionalGeneration,
-    AutoModelForCausalLM
-)
-import torch
-from peft import PeftModel
-from datasets import load_from_disk
-import pandas as pd
-from tqdm import tqdm
-from torch.utils.data import DataLoader
-
-
-mode_path = '/gemini/pretrain/meta-llamaLlama-3.2-11B-Vision-Instruct'
-lora_path = '/gemini/code/FMD/model/final_model_4/checkpoint-2440' # lora 输出对应 checkpoint 路径
-
-# 加载tokenizer
-tokenizer = AutoTokenizer.from_pretrained(mode_path, trust_remote_code=True)
-
-# 加载模型
-model = MllamaForConditionalGeneration.from_pretrained(mode_path, device_map="auto",torch_dtype=torch.bfloat16, trust_remote_code=True).eval()
-
-# 加载lora权重
-model = PeftModel.from_pretrained(model, model_id=lora_path)
-test_dataset = load_from_disk("/gemini/code/FMD/final_dataset/Test")
-results = []
-with torch.no_grad():
-    for data in tqdm(test_dataset):
-        model_input = tokenizer(
-            data['instruction_1'],                  # 输入文本
-            add_special_tokens=False,     # 不添加特殊标记
-            truncation=True,              # 启用截断
-            max_length=3000              # 设置最大长度
-        )
-        model_input = tokenizer.decode(model_input["input_ids"], skip_special_tokens=False)
-        
-        model_inputs = tokenizer(f"<|begin_of_text|><|start_header_id|>system<|end_header_id|>\n\nYou are an expert in financial misinformation detection.<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{model_input}\nimage information: {data['image_info']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n", truncation=True, max_length=3600, add_special_tokens=False,return_tensors="pt").to('cuda')
-        # 生成模型输出
-        generated_ids = model.generate(**model_inputs, max_new_tokens=1024)
-        
-        # 去除输入部分的 token，以保留生成的预测结果
-        generated_ids = [
-            output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
-        ]
-        
-        # 解码生成的预测结果
-        responses = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
-        print(responses)
-        # 将每个结果按顺序存储到列表中
-        results.append({
-            "ID": data['ID'],
-            "response": responses
-        })
-if results:
-    df = pd.DataFrame(results)
-    output_csv = "/gemini/code/FMD/inference/result_final_model_4/response.csv"
-    df.to_csv(output_csv, index=False, encoding='utf-8')
-    print(f"Results saved to {output_csv}")