Update app.py

app.py

@@ -1,61 +1,87 @@
+import os
 import gradio as gr
-from transformers import AutoModel, AutoTokenizer
+import copy
+from llama_cpp import Llama
+from huggingface_hub import hf_hub_download  
 
-# Load the model
-model_name = "wop/kosmox-gguf"
-model = AutoModel.from_pretrained(model_name)
 
-# Assuming we need to load a corresponding tokenizer
-tokenizer = AutoTokenizer.from_pretrained(model_name)
+llm = Llama(
+    model_path=hf_hub_download(
+        repo_id=os.environ.get("REPO_ID", "microsoft/Phi-3-mini-4k-instruct-gguf"),
+        filename=os.environ.get("MODEL_FILE", "Phi-3-mini-4k-instruct-q4.gguf"),
+    ),
+    n_ctx=2048,
+    n_gpu_layers=50, # change n_gpu_layers if you have more or less VRAM 
+) 
 
-# Define the chat template
-chat_template = "{{ bos_token }}{% for message in messages %}{% if message['from'] == 'human' %}{{' ' + message['value'] + ' '}}{% elif message['from'] == 'gpt' %}{{' ' + message['value'] + ' '}}{% else %}{{'<|' + message['from'] + '|> ' + message['value'] + ' '}}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ ' ' }}{% endif %}"
 
-def chat(messages):
-    # Prepare the input
-    context = chat_template.format(
-        bos_token=tokenizer.bos_token,
-        messages=messages,
-        add_generation_prompt=True
+def generate_text(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+):
+    temp = ""
+    input_prompt = f"[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n "
+    for interaction in history:
+        input_prompt = input_prompt + str(interaction[0]) + " [/INST] " + str(interaction[1]) + " </s><s> [INST] "
+
+    input_prompt = input_prompt + str(message) + " [/INST] "
+
+    output = llm(
+        input_prompt,
+        temperature=temperature,
+        top_p=top_p,
+        top_k=40,
+        repeat_penalty=1.1,
+        max_tokens=max_tokens,
+        stop=[
+            "<|prompter|>",
+            "<|endoftext|>",
+            "<|endoftext|> \n",
+            "ASSISTANT:",
+            "USER:",
+            "SYSTEM:",
+        ],
+        stream=True,
     )
-    
-    # Tokenize the input
-    inputs = tokenizer(context, return_tensors='pt')
-    
-    # Generate response
-    outputs = model.generate(**inputs)
-    
-    # Decode the response
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    
-    return response
-
-# Define the Gradio interface
-with gr.Blocks() as demo:
-    chatbot = gr.Chatbot()
-    with gr.Row():
-        with gr.Column():
-            user_input = gr.Textbox(
-                placeholder="Type your message here...",
-                label="Your message"
-            )
-            send_button = gr.Button("Send")
-        with gr.Column():
-            chat_output = gr.Textbox(
-                label="Chatbot response",
-                interactive=False
-            )
-
-    def respond(message, history):
-        history = history or []
-        history.append({"from": "human", "value": message})
-        
-        response = chat(history)
-        history.append({"from": "gpt", "value": response})
-        
-        return history, history[-1]['value']
-
-    send_button.click(respond, [user_input, chatbot], [chatbot, chat_output])
-
-# Launch the Gradio interface
-demo.launch()
+    for out in output:
+        stream = copy.deepcopy(out)
+        temp += stream["choices"][0]["text"]
+        yield temp
+
+
+demo = gr.ChatInterface(
+    generate_text,
+    title="llama-cpp-python on GPU",
+    description="Running LLM with https://github.com/abetlen/llama-cpp-python",
+    examples=[
+        ['How to setup a human base on Mars? Give short answer.'],
+        ['Explain theory of relativity to me like I’m 8 years old.'],
+        ['What is 9,000 * 9,000?'],
+        ['Write a pun-filled happy birthday message to my friend Alex.'],
+        ['Justify why a penguin might make a good king of the jungle.']
+    ],
+    cache_examples=False,
+    retry_btn=None,
+    undo_btn="Delete Previous",
+    clear_btn="Clear",
+    additional_inputs=[
+        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
+        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
+        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
+        gr.Slider(
+            minimum=0.1,
+            maximum=1.0,
+            value=0.95,
+            step=0.05,
+            label="Top-p (nucleus sampling)",
+        ),
+    ],
+)
+
+
+if __name__ == "__main__":
+    demo.launch()