Fix LLM with LLama3

app.py

@@ -3,7 +3,7 @@ from huggingface_hub import InferenceClient
 
 """
 For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
+
 client = InferenceClient("meta-llama/Meta-Llama-3-8B")
 
 
@@ -38,13 +38,42 @@ def respond(
 
         response += token
         yield response
+"""
+model_id = "meta-llama/Meta-Llama-3-8B-Instruct"
+
+pipeline = transformers.pipeline(
+    "text-generation",
+    model=model_id,
+    model_kwargs={"torch_dtype": torch.bfloat16},
+    device="cuda",
+)
+
+def chat_function(message, history, system_prompt, max_new_tokens, temperature):
+    messages = [{"role":"system","content":system_prompt},
+                {"role":"user", "content":message}]
+    prompt = pipeline.tokenizer.apply_chat_template(
+        messages,
+        tokenize=False,
+        add_generation_prompt=True,)
+    terminators = [
+        pipeline.tokenizer.eos_token_id,
+        pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")]
+    outputs = pipeline(
+        prompt,
+        max_new_tokens = max_new_tokens,
+        eos_token_id = terminators,
+        do_sample = True,
+        temperature = temperature + 0.1,
+        top_p = 0.9,)
+    return outputs[0]["generated_text"][len(prompt):]
+
 
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
 """
             
 demo = gr.ChatInterface(
-    respond,
+    chat_function,
     textbox=gr.Textbox(placeholder="Enter message here", container=False, scale = 7),
     chatbot=gr.Chatbot(height=400),
     additional_inputs=[