Update app.py

app.py

@@ -1,13 +1,19 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer
+import requests
 
-model = AutoModelForCausalLM.from_pretrained("mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated", device_map="auto")
-tokenizer = AutoTokenizer.from_pretrained("mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated")
+API_URL = "https://api-inference.huggingface.co/models/mlabonne/Meta-Llama-3.1-8B-Instruct-abliterated"
+headers = {"Authorization": "Bearer hf_..."}  # Replace with your Hugging Face API token
+
+def query(payload):
+    response = requests.post(API_URL, headers=headers, json=payload)
+    return response.json()
 
 def generate_text(prompt):
-    input_ids = tokenizer.encode(prompt, return_tensors="pt")
-    output = model.generate(input_ids, max_length=200)
-    return tokenizer.decode(output[0], skip_special_tokens=True)
+    output = query({
+        "inputs": prompt,
+        "parameters": {"max_new_tokens": 200}
+    })
+    return output[0]['generated_text']
 
 iface = gr.Interface(fn=generate_text, inputs="text", outputs="text")
 iface.launch()