Update app.py

app.py

@@ -1,168 +1,149 @@
-import os
-import time
 import spaces
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
+import os
+import subprocess
+from llama_cpp import Llama
+from llama_cpp_agent import LlamaCppAgent, MessagesFormatterType
+from llama_cpp_agent.providers import LlamaCppPythonProvider
+from llama_cpp_agent.chat_history import BasicChatHistory
+from llama_cpp_agent.chat_history.messages import Roles
 import gradio as gr
-from threading import Thread
-
-MODEL_LIST = ["meta-llama/Meta-Llama-3.1-8B-Instruct"]
-HF_TOKEN = os.environ.get("HF_TOKEN", None)
-MODEL = os.environ.get("MODEL_ID")
-
-TITLE = "<h1><center>Meta-Llama3.1-8B</center></h1>"
-
-PLACEHOLDER = """
-<center>
-<p>Hi! How can I help you today?</p>
-</center>
-"""
-
-
-CSS = """
-.duplicate-button {
-    margin: auto !important;
-    color: white !important;
-    background: black !important;
-    border-radius: 100vh !important;
-}
-h3 {
-    text-align: center;
-}
-"""
-
-device = "cuda" # for GPU usage or "cpu" for CPU usage
-
-quantization_config = BitsAndBytesConfig(
-    load_in_4bit=True,
-    bnb_4bit_compute_dtype=torch.bfloat16,
-    bnb_4bit_use_double_quant=True,
-    bnb_4bit_quant_type= "nf4")
-
-tokenizer = AutoTokenizer.from_pretrained(MODEL)
-model = AutoModelForCausalLM.from_pretrained(
-    MODEL,
-    torch_dtype=torch.bfloat16,
-    device_map="auto",
-    quantization_config=quantization_config)
-
-@spaces.GPU()
-def stream_chat(
-    message: str, 
-    history: list,
-    system_prompt: str,
-    temperature: float = 0.8, 
-    max_new_tokens: int = 1024, 
-    top_p: float = 1.0, 
-    top_k: int = 20, 
-    penalty: float = 1.2,
+from huggingface_hub import hf_hub_download
+
+huggingface_token = os.getenv("HF_TOKEN")
+
+# Download the Meta-Llama-3.1-8B-Instruct model
+hf_hub_download(
+    repo_id="itaudit/2024-08-14_V3_394_sample",
+    filename="unsloth.Q4_K_M.gguf",
+    local_dir="./models",
+    token=huggingface_token
+)
+
+llm = None
+llm_model = None
+
+@spaces.GPU(duration=120)
+def respond(
+    message,
+    history: list[tuple[str, str]],
+    model,
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    top_k,
+    repeat_penalty,
 ):
-    print(f'message: {message}')
-    print(f'history: {history}')
-
-    conversation = [
-        {"role": "system", "content": system_prompt}
-    ]
-    for prompt, answer in history:
-        conversation.extend([
-            {"role": "user", "content": prompt}, 
-            {"role": "assistant", "content": answer},
-        ])
-
-    conversation.append({"role": "user", "content": message})
+    chat_template = MessagesFormatterType.GEMMA_2
 
-    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt").to(model.device)
+    global llm
+    global llm_model
     
-    streamer = TextIteratorStreamer(tokenizer, timeout=60.0, skip_prompt=True, skip_special_tokens=True)
-    
-    generate_kwargs = dict(
-        input_ids=input_ids, 
-        max_new_tokens = max_new_tokens,
-        do_sample = False if temperature == 0 else True,
-        top_p = top_p,
-        top_k = top_k,
-        temperature = temperature,
-        repetition_penalty=penalty,
-        eos_token_id=[128001,128008,128009],
-        streamer=streamer,
+    # Load model only if it's not already loaded or if a new model is selected
+    if llm is None or llm_model != model:
+        try:
+            llm = Llama(
+                model_path=f"models/{model}",
+                flash_attn=True,
+                n_gpu_layers=81,  # Adjust based on available GPU resources
+                n_batch=1024,
+                n_ctx=8192,
+            )
+            llm_model = model
+        except Exception as e:
+            return f"Error loading model: {str(e)}"
+
+    provider = LlamaCppPythonProvider(llm)
+
+    agent = LlamaCppAgent(
+        provider,
+        system_prompt=f"{system_message}",
+        predefined_messages_formatter_type=chat_template,
+        debug_output=True
     )
-
-    with torch.no_grad():
-        thread = Thread(target=model.generate, kwargs=generate_kwargs)
-        thread.start()
+    
+    settings = provider.get_provider_default_settings()
+    settings.temperature = temperature
+    settings.top_k = top_k
+    settings.top_p = top_p
+    settings.max_tokens = max_tokens
+    settings.repeat_penalty = repeat_penalty
+    settings.stream = True
+
+    messages = BasicChatHistory()
+
+    # Add user and assistant messages to the history
+    for msn in history:
+        user = {'role': Roles.user, 'content': msn[0]}
+        assistant = {'role': Roles.assistant, 'content': msn[1]}
+        messages.add_message(user)
+        messages.add_message(assistant)
+    
+    # Stream the response
+    try:
+        stream = agent.get_chat_response(
+            message,
+            llm_sampling_settings=settings,
+            chat_history=messages,
+            returns_streaming_generator=True,
+            print_output=False
+        )
         
-    buffer = ""
-    for new_text in streamer:
-        buffer += new_text
-        yield buffer
-
-            
-chatbot = gr.Chatbot(height=600, placeholder=PLACEHOLDER)
-
-with gr.Blocks(css=CSS, theme="soft") as demo:
-    gr.HTML(TITLE)
-    gr.DuplicateButton(value="Duplicate Space for private use", elem_classes="duplicate-button")
-    gr.ChatInterface(
-        fn=stream_chat,
-        chatbot=chatbot,
-        fill_height=True,
-        additional_inputs_accordion=gr.Accordion(label="⚙️ Parameters", open=False, render=False),
-        additional_inputs=[
-            gr.Textbox(
-                value="You are a helpful assistant",
-                label="System Prompt",
-                render=False,
-            ),
-            gr.Slider(
-                minimum=0,
-                maximum=1,
-                step=0.1,
-                value=0.8,
-                label="Temperature",
-                render=False,
-            ),
-            gr.Slider(
-                minimum=128,
-                maximum=8192,
-                step=1,
-                value=1024,
-                label="Max new tokens",
-                render=False,
-            ),
-            gr.Slider(
-                minimum=0.0,
-                maximum=1.0,
-                step=0.1,
-                value=1.0,
-                label="top_p",
-                render=False,
-            ),
-            gr.Slider(
-                minimum=1,
-                maximum=20,
-                step=1,
-                value=20,
-                label="top_k",
-                render=False,
-            ),
-            gr.Slider(
-                minimum=0.0,
-                maximum=2.0,
-                step=0.1,
-                value=1.2,
-                label="Repetition penalty",
-                render=False,
-            ),
-        ],
-        examples=[
-            ["Help me study vocabulary: write a sentence for me to fill in the blank, and I'll try to pick the correct option."],
-            ["What are 5 creative things I could do with my kids' art? I don't want to throw them away, but it's also so much clutter."],
-            ["Tell me a random fun fact about the Roman Empire."],
-            ["Show me a code snippet of a website's sticky header in CSS and JavaScript."],
-        ],
-        cache_examples=False,
+        outputs = ""
+        for output in stream:
+            outputs += output
+            yield outputs
+    except Exception as e:
+        yield f"Error during response generation: {str(e)}"
+
+description = """<p align="center">Using the Meta-Llama-3.1-8B-Instruct Model</p>"""
+
+demo = gr.ChatInterface(
+    respond,
+    additional_inputs=[
+        gr.Dropdown([
+                'Meta-Llama-3.1-8B-Instruct-Q5_K_M.gguf'
+            ],
+            value="Meta-Llama-3.1-8B-Instruct-Q5_K_M.gguf",
+            label="Model"
+        ),
+        gr.Textbox(value="You are a helpful assistant.", label="System message"),
+        gr.Slider(minimum=1, maximum=4096, value=2048, step=1, label="Max 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",
+        ),
+        gr.Slider(
+            minimum=0,
+            maximum=100,
+            value=40,
+            step=1,
+            label="Top-k",
+        ),
+        gr.Slider(
+            minimum=0.0,
+            maximum=2.0,
+            value=1.1,
+            step=0.1,
+            label="Repetition penalty",
+        ),
+    ],
+    retry_btn="Retry",
+    undo_btn="Undo",
+    clear_btn="Clear",
+    submit_btn="Send",
+    title="Chat with Meta-Llama-3.1-8B-Instruct using llama.cpp", 
+    description=description,
+    chatbot=gr.Chatbot(
+        scale=1, 
+        likeable=False,
+        show_copy_button=True
     )
-
+)
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()