Update app.py

app.py

@@ -1,68 +1,105 @@
-import streamlit as st
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
-import random
+import os
+os.system('pip install dashscope')
+import gradio as gr
+from http import HTTPStatus
+import dashscope
+from dashscope import Generation
+from dashscope.api_entities.dashscope_response import Role
+from typing import List, Optional, Tuple, Dict
+from urllib.error import HTTPError
+default_system = 'You are a helpful assistant.'
 
-@st.cache_resource
-def load_model():
-    model_name = "sberbank-ai/rugpt3small_based_on_gpt2"
-    tokenizer = AutoTokenizer.from_pretrained(model_name)
-    model = AutoModelForCausalLM.from_pretrained(model_name)
-    return tokenizer, model
+YOUR_API_TOKEN = os.getenv('YOUR_API_TOKEN')
+dashscope.api_key = YOUR_API_TOKEN
 
-def generate_response(prompt, tokenizer, model):
-    inputs = tokenizer.encode(prompt, return_tensors='pt')
-    with torch.no_grad():
-        outputs = model.generate(inputs, max_length=100, num_return_sequences=1, 
-                                 temperature=0.9, top_k=50, top_p=0.95)
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return add_mistakes(response)
+History = List[Tuple[str, str]]
+Messages = List[Dict[str, str]]
 
-def add_mistakes(text):
-    words = text.split()
-    for i in range(len(words)):
-        if random.random() < 0.2:  # 20% шанс ошибки в слове
-            words[i] = misspell_word(words[i])
-    return ' '.join(words)
+def clear_session() -> History:
+    return '', []
 
-def misspell_word(word):
-    if len(word) < 3:
-        return word
-    vowels = 'аеёиоуыэюя'
-    consonants = 'бвгджзйклмнпрстфхцчшщ'
-    
-    if random.random() < 0.5:
-        # Заменяем случайную гласную
-        for i, char in enumerate(word):
-            if char.lower() in vowels:
-                replacement = random.choice(vowels)
-                return word[:i] + replacement + word[i+1:]
-    else:
-        # Заменяем случайную согласную
-        for i, char in enumerate(word):
-            if char.lower() in consonants:
-                replacement = random.choice(consonants)
-                return word[:i] + replacement + word[i+1:]
-    return word
+def modify_system_session(system: str) -> str:
+    if system is None or len(system) == 0:
+        system = default_system
+    return system, system, []
 
-st.title("AI Чат с простой русской моделью")
+def history_to_messages(history: History, system: str) -> Messages:
+    messages = [{'role': Role.SYSTEM, 'content': system}]
+    for h in history:
+        messages.append({'role': Role.USER, 'content': h[0]})
+        messages.append({'role': Role.ASSISTANT, 'content': h[1]})
+    return messages
 
-tokenizer, model = load_model()
 
-if "messages" not in st.session_state:
-    st.session_state.messages = []
+def messages_to_history(messages: Messages) -> Tuple[str, History]:
+    assert messages[0]['role'] == Role.SYSTEM
+    system = messages[0]['content']
+    history = []
+    for q, r in zip(messages[1::2], messages[2::2]):
+        history.append([q['content'], r['content']])
+    return system, history
 
-for message in st.session_state.messages:
-    with st.chat_message(message["role"]):
-        st.markdown(message["content"])
 
-if prompt := st.chat_input("Введите ваше сообщение"):
-    st.session_state.messages.append({"role": "user", "content": prompt})
-    with st.chat_message("user"):
-        st.markdown(prompt)
-    
-    with st.chat_message("assistant"):
-        response = generate_response(prompt, tokenizer, model)
-        st.markdown(response)
+def model_chat(query: Optional[str], history: Optional[History], system: str
+) -> Tuple[str, str, History]:
+    if query is None:
+        query = ''
+    if history is None:
+        history = []
+    messages = history_to_messages(history, system)
+    messages.append({'role': Role.USER, 'content': query})
+    gen = Generation.call(
+        model='qwen2-72b-instruct',
+        messages=messages,
+        result_format='message',
+        stream=True
+    )
+    for response in gen:
+        if response.status_code == HTTPStatus.OK:
+            role = response.output.choices[0].message.role
+            response = response.output.choices[0].message.content
+            system, history = messages_to_history(messages + [{'role': role, 'content': response}])
+            yield '', history, system
+        else:
+            raise ValueError('Request id: %s, Status code: %s, error code: %s, error message: %s' % (
+                response.request_id, response.status_code,
+                response.code, response.message
+            ))
+
+
+with gr.Blocks() as demo:
+    gr.Markdown("""<center><font size=8>Qwen2-72B-instruct Chat👾</center>""")
+
+    with gr.Row():
+        with gr.Column(scale=3):
+            system_input = gr.Textbox(value=default_system, lines=1, label='System')
+        with gr.Column(scale=1):
+            modify_system = gr.Button("🛠️ Set system prompt and clear history", scale=2)
+        system_state = gr.Textbox(value=default_system, visible=False)
+    chatbot = gr.Chatbot(label='qwen2-72B-instruct')
+    textbox = gr.Textbox(lines=1, label='Input')
+
+    with gr.Row():
+        clear_history = gr.Button("🧹 Clear history")
+        sumbit = gr.Button("🚀 Send")
+
+    textbox.submit(model_chat,
+                 inputs=[textbox, chatbot, system_state],
+                 outputs=[textbox, chatbot, system_input],
+                 concurrency_limit = 40)
     
-    st.session_state.messages.append({"role": "assistant", "content": response})
+    sumbit.click(model_chat,
+                 inputs=[textbox, chatbot, system_state],
+                 outputs=[textbox, chatbot, system_input],
+                 concurrency_limit = 40)
+    clear_history.click(fn=clear_session,
+                        inputs=[],
+                        outputs=[textbox, chatbot],
+                        concurrency_limit = 40)
+    modify_system.click(fn=modify_system_session,
+                        inputs=[system_input],
+                        outputs=[system_state, system_input, chatbot],
+                        concurrency_limit = 40)
+
+demo.queue(api_open=False)
+demo.launch(max_threads=40)