try llama 3

app.py

@@ -2,7 +2,7 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 import os
 from transformers import AutoTokenizer, AutoModelForCausalLM
-
+import requests
 
 # ACCESS_TOKEN = os.getenv('ACCESS_TOKEN')
 """
@@ -48,10 +48,16 @@ def respond(
     #     response += token
     #     yield response
 
-    input_ids = tokenizer.encode(message, return_tensors = 'pt')
-    for output in model.generate(input_ids, stream=True):
-        output_text = tokenizer.decode(output, skip_special_tokens=True)
-        yield output_text
+    ### doesn't work
+    # input_ids = tokenizer.encode(message, return_tensors = 'pt')
+    # for output in model.generate(input_ids, stream=True):
+    #     output_text = tokenizer.decode(output, skip_special_tokens=True)
+    #     yield output_text
+    
+    API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-8B-Instruct"
+    headers = {"Authorization": "Bearer "+os.environ['hf_token']}
+    response = requests.post(API_URL, headers=headers, json={"inputs":"message"})
+    return response.json()