Fix 8

app.py

@@ -1,26 +1,38 @@
 import gradio as gr
-from openai import OpenAI
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+import torch
 import os
 from dotenv import load_dotenv
+from huggingface_hub import login
 
 load_dotenv()
 
-# Initialiser le client
-client = OpenAI(
-    base_url="https://api-inference.huggingface.co/v1",
-    api_key=os.getenv('HF_TOKEN')
+# Login to Hugging Face
+hf_token = os.getenv('HF_TOKEN')
+login(hf_token)
+
+# Configuration du modèle
+model_path = "Nac31/Sacha-Mistral-0"
+dtype = torch.bfloat16 if torch.cuda.get_device_capability()[0] == 8 else torch.float16
+
+# Initialisation du modèle
+tokenizer = AutoTokenizer.from_pretrained(model_path)
+model = AutoModelForCausalLM.from_pretrained(
+    model_path,
+    device_map="auto",
+    torch_dtype=dtype
 )
+pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
 def generate_response(message, temperature=0.7):
     try:
-        response = client.chat.completions.create(
-            model="Nac31/Sacha-Mistral-0",
-            messages=[{"role": "user", "content": message}],
+        response = pipe(
+            message,
             temperature=temperature,
-            max_tokens=500,
-            stream=False
+            max_new_tokens=500,
+            do_sample=True
         )
-        return response.choices[0].message.content
+        return response[0]['generated_text']
     except Exception as e:
         return f"Une erreur s'est produite : {str(e)}"