Update app.py
Browse files
app.py
CHANGED
|
@@ -1,86 +1,23 @@
|
|
| 1 |
-
import gradio as gr
|
| 2 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 3 |
import torch
|
| 4 |
|
| 5 |
-
#
|
| 6 |
-
model_path = "tosei0000/code-AI" #
|
| 7 |
|
| 8 |
-
#
|
| 9 |
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
|
| 10 |
-
model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True
|
| 11 |
-
model.eval()
|
| 12 |
|
| 13 |
-
#
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
"也能像朋友一样聊天。请保持耐心、有趣,尽可能详细地回答问题。\n"
|
| 17 |
-
)
|
| 18 |
|
| 19 |
-
#
|
| 20 |
-
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
-
#
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
# 更新历史上下文
|
| 27 |
-
chat_history.append(f"用户: {user_input}")
|
| 28 |
-
if len(chat_history) > 5:
|
| 29 |
-
chat_history = chat_history[-5:] # 只保留最近 5 条
|
| 30 |
-
|
| 31 |
-
# 拼接 prompt
|
| 32 |
-
if chat_mode == "代码生成":
|
| 33 |
-
prompt = f"{SYSTEM_PROMPT}\n请根据以下需求生成代码:\n{user_input}\n"
|
| 34 |
-
else: # 聊天模式
|
| 35 |
-
prompt = SYSTEM_PROMPT + "\n" + "\n".join(chat_history) + "\n助手:"
|
| 36 |
-
|
| 37 |
-
inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=4096)
|
| 38 |
-
with torch.no_grad():
|
| 39 |
-
outputs = model.generate(
|
| 40 |
-
**inputs,
|
| 41 |
-
max_new_tokens=max_tokens,
|
| 42 |
-
temperature=temperature,
|
| 43 |
-
top_p=0.95,
|
| 44 |
-
do_sample=True,
|
| 45 |
-
eos_token_id=tokenizer.eos_token_id,
|
| 46 |
-
)
|
| 47 |
-
|
| 48 |
-
decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 49 |
-
|
| 50 |
-
# 提取最后一句助手的回复
|
| 51 |
-
reply = decoded_output.split("助手:")[-1].strip()
|
| 52 |
-
|
| 53 |
-
# 保存助手回复到历史中
|
| 54 |
-
chat_history.append(f"助手: {reply}")
|
| 55 |
-
|
| 56 |
-
return reply
|
| 57 |
-
|
| 58 |
-
# 重置历史按钮
|
| 59 |
-
def reset_memory():
|
| 60 |
-
global chat_history
|
| 61 |
-
chat_history = []
|
| 62 |
-
return "记忆已重置。"
|
| 63 |
-
|
| 64 |
-
# Gradio 界面
|
| 65 |
-
with gr.Blocks() as demo:
|
| 66 |
-
gr.Markdown("# 🤖 智能代码助理 + 聊天机器人")
|
| 67 |
-
gr.Markdown("支持代码生成与聊天功能,可记忆上下文,具备人格设定!")
|
| 68 |
-
|
| 69 |
-
with gr.Row():
|
| 70 |
-
chat_mode = gr.Radio(["聊天", "代码生成"], value="代码生成", label="对话模式")
|
| 71 |
-
reset_btn = gr.Button("🧹 重置记忆")
|
| 72 |
-
|
| 73 |
-
user_input = gr.Textbox(label="你的输入", lines=6, placeholder="输入代码需求或聊天内容...")
|
| 74 |
-
max_tokens = gr.Slider(50, 1024, value=512, step=10, label="最大生成长度")
|
| 75 |
-
temperature = gr.Slider(0.1, 1.5, value=0.7, step=0.1, label="生成多样性(temperature)")
|
| 76 |
-
|
| 77 |
-
output = gr.Textbox(label="AI 回复", lines=10)
|
| 78 |
-
|
| 79 |
-
submit_btn = gr.Button("✨ 生成")
|
| 80 |
-
|
| 81 |
-
submit_btn.click(fn=generate_reply, inputs=[user_input, chat_mode, max_tokens, temperature], outputs=output)
|
| 82 |
-
reset_btn.click(fn=reset_memory, outputs=output)
|
| 83 |
-
|
| 84 |
-
# 启动服务
|
| 85 |
-
if __name__ == "__main__":
|
| 86 |
-
demo.launch()
|
|
|
|
|
|
|
| 1 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 2 |
import torch
|
| 3 |
|
| 4 |
+
# 设置模型路径(替换为你上传后显示的路径名)
|
| 5 |
+
model_path = "tosei0000/code-AI" # 修改为你上传的文件夹名
|
| 6 |
|
| 7 |
+
# 加载模型和分词器
|
| 8 |
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
|
| 9 |
+
model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True)
|
|
|
|
| 10 |
|
| 11 |
+
# 如果你有 GPU(Kaggle 支持 GPU),把模型放到 GPU
|
| 12 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 13 |
+
model.to(device)
|
|
|
|
|
|
|
| 14 |
|
| 15 |
+
# 推理函数
|
| 16 |
+
def chat(prompt, max_new_tokens=100):
|
| 17 |
+
inputs = tokenizer(prompt, return_tensors="pt").to(device)
|
| 18 |
+
outputs = model.generate(**inputs, max_new_tokens=max_new_tokens)
|
| 19 |
+
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 20 |
|
| 21 |
+
# 测试
|
| 22 |
+
response = chat("你好,请介绍一下你自己。")
|
| 23 |
+
print(response)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|