Upload 2 files

app.py

@@ -1,14 +1,14 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoModelForCausalLM, AutoTokenizer
 import torch
 
-# 加载轻量模型（CPU 兼容）
-tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/deepseek-coder-1.3b-base")
-model = AutoModelForCausalLM.from_pretrained("deepseek-ai/deepseek-coder-1.3b-base")
+# 加载模型
+model_name = "deepseek-ai/deepseek-coder-1.3b-base"  # 可替换为 "deepseek-ai/deepseek-coder-1.3b-instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
     model_name,
     torch_dtype=torch.float16,  # 使用 FP16 减少内存
-    device_map="cpu",          # 强制使用 CPU
+    device_map="cpu",          # 强制 CPU
     trust_remote_code=True,
     low_cpu_mem_usage=True
 )
@@ -31,8 +31,13 @@ def respond(
 
     messages.append({"role": "user", "content": message})
 
-    # 使用聊天模板格式化输入
-    input_text = tokenizer.apply_chat_template(messages, tokenize=False)
+    # 使用聊天模板格式化输入（base 模型可能无模板，需调整）
+    try:
+        input_text = tokenizer.apply_chat_template(messages, tokenize=False)
+    except:
+        # 如果 base 模型无聊天模板，直接拼接
+        input_text = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages])
+
     inputs = tokenizer(input_text, return_tensors="pt").to("cpu")
 
     # 生成响应
@@ -52,7 +57,7 @@ demo = gr.ChatInterface(
     respond,
     additional_inputs=[
         gr.Textbox(value="You are a friendly coding assistant.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=1, maximum=2048, value=256, step=1, label="Max new tokens"),  # 降低以加快 CPU 推理
         gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
         gr.Slider(
             minimum=0.1,

requirements.txt

@@ -1,5 +1,4 @@
 torch>=2.0.0
 transformers>=4.38.0
 accelerate>=0.26.0
-gradio>=4.0.0
-bitsandbytes>=0.43.3
+gradio>=4.0.0