fix

app.py

@@ -1,65 +1,39 @@
 import gradio as gr
 import spaces
 import torch
-from transformers import LlamaTokenizerFast, LlamaForCausalLM
 from peft import PeftModel
 import os
+from datasets import load_dataset
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
-# 加载模型和tokenizer
-model_name = "meta-llama/Meta-Llama-3-8B"
-adapter_name = "FinGPT/fingpt-mt_llama3-8b_lora"
 
 # 获取HF token（Spaces会自动提供）
 hf_token = os.environ.get("HF_TOKEN") or os.environ.get("HUGGING_FACE_HUB_TOKEN")
 
-print("=" * 50)
-print("开始初始化模型...")
-print(f"Token存在: {bool(hf_token)}")
-print(f"基础模型: {model_name}")
-print(f"LoRA适配器: {adapter_name}")
-print("=" * 50)
 
 model = None
 tokenizer = None
-model_loaded = False
 
 try:
-    print("\n[1/3] 加载tokenizer...")
-    tokenizer = LlamaTokenizerFast.from_pretrained(
-        model_name,
-        trust_remote_code=True,
-        token=hf_token,
-    )
-    tokenizer.pad_token = tokenizer.eos_token
-    print("✓ Tokenizer加载成功")
 
-    print("\n[2/3] 加载基础模型...")
-    base_model = LlamaForCausalLM.from_pretrained(
-        model_name,
-        torch_dtype=torch.float16,
-        device_map="auto",
+    base_model = AutoModelForCausalLM.from_pretrained(
+        'meta-llama/Llama-2-7b-chat-hf',
         trust_remote_code=True,
+        device_map="auto",
+        torch_dtype=torch.float16,   # optional if you have enough VRAM
         token=hf_token,
     )
-    print("\n[3/3] 加载LoRA适配器...")
-    model = PeftModel.from_pretrained(base_model, adapter_name)
+    tokenizer = AutoTokenizer.from_pretrained('meta-llama/Llama-2-7b-chat-hf',
+                                              trust_remote_code=True,
+                                              token=hf_token,
+                                              )
+    model = PeftModel.from_pretrained(base_model, 'FinGPT/fingpt-forecaster_dow30_llama2-7b_lora')
     model = model.eval()
 
-    print("✓ LoRA适配器加载成功")
-    print("\n" + "=" * 50)
-    print("✅ 所有模型组件加载完成！")
-    print("=" * 50 + "\n")
-
 except Exception as e:
     print("\n" + "=" * 50)
     print("❌ 模型加载失败！")
     print(f"错误信息: {e}")
-    print("\n可能的解决方案:")
-    print("1. 确保你的HF账号已获得Meta-Llama-3-8B访问权限")
-    print("   访问: https://huggingface.co/meta-llama/Meta-Llama-3-8B")
-    print("2. 在Spaces的Settings → Repository secrets中添加HF_TOKEN")
-    print("   或启用 'hf_oauth: true' 以使用OAuth认证")
-    print("=" * 50 + "\n")
     raise
 
 
@@ -116,7 +90,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
         """
         # 🤖 FinGPT Chatbot
         
-        这是一个基于 **FinGPT/fingpt-mt_llama3-8b_lora** 模型的金融对话助手。
+        这是一个基于 **FinGPT/fingpt-forecaster_dow30_llama2-7b_lora** 模型的金融对话助手。
         
         您可以询问关于金融市场、投资、经济分析等问题。
         """