Fix ?

app.py

@@ -1,25 +1,24 @@
 import gradio as gr
-from transformers import AutoTokenizer, pipeline
+from huggingface_hub import InferenceClient
+import os
+from dotenv import load_dotenv
 
-def initialize_model():
-    tokenizer = AutoTokenizer.from_pretrained("Nac31/Sacha-Mistral-0", trust_remote_code=True)
-    pipe = pipeline(
-        "text-generation",
-        model="Nac31/Sacha-Mistral-0",
-        tokenizer=tokenizer,
-        trust_remote_code=True
-    )
-    return pipe
+load_dotenv()
 
-def generate_response(message, temperature=0.7, max_length=500):
+# Initialiser le client
+client = InferenceClient(
+    model="Nac31/Sacha-Mistral-0",
+    token=os.getenv("HF_TOKEN")  # Votre token HF
+)
+
+def generate_response(message, temperature=0.7):
     try:
-        pipe = initialize_model()
-        response = pipe(
+        response = client.text_generation(
             message,
-            max_length=max_length,
+            max_new_tokens=500,
             temperature=temperature,
             do_sample=True
-        )[0]['generated_text']
+        )
         return response
     except Exception as e:
         return f"Une erreur s'est produite : {str(e)}"

requirements.txt

@@ -9,4 +9,5 @@ bitsandbytes
 openai
 langchain
 python-dotenv
-langchain-community
+langchain-community
+huggingface_hub