Spaces:
Runtime error
Runtime error
Create main.py
Browse files
main.py
ADDED
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
3 |
+
import torch
|
4 |
+
|
5 |
+
# Charger le modèle
|
6 |
+
model_name = "mistralai/Mistral-7B-Instruct-v0.2"
|
7 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
8 |
+
model = AutoModelForCausalLM.from_pretrained(
|
9 |
+
model_name,
|
10 |
+
torch_dtype=torch.float16,
|
11 |
+
device_map="auto"
|
12 |
+
)
|
13 |
+
|
14 |
+
# Prompt système pour personnaliser Aria
|
15 |
+
system_prompt = """Tu es Aria, une IA bienveillante et polie qui répond de façon concise et claire."""
|
16 |
+
|
17 |
+
def chat(message, history=[]):
|
18 |
+
prompt = system_prompt + "\n" + "\n".join([f"Utilisateur: {m[0]}\nAria: {m[1]}" for m in history]) + f"\nUtilisateur: {message}\nAria:"
|
19 |
+
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
20 |
+
outputs = model.generate(**inputs, max_new_tokens=200)
|
21 |
+
reply = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
22 |
+
# Récupérer uniquement la réponse d'Aria
|
23 |
+
reply = reply.split("Aria:")[-1].strip()
|
24 |
+
history.append((message, reply))
|
25 |
+
return reply, history
|
26 |
+
|
27 |
+
# Interface Gradio
|
28 |
+
with gr.Blocks() as demo:
|
29 |
+
chatbot = gr.Chatbot()
|
30 |
+
msg = gr.Textbox(placeholder="Écris un message...")
|
31 |
+
msg.submit(chat, [msg, chatbot], [chatbot, chatbot])
|
32 |
+
|
33 |
+
demo.launch()
|