Create app.py

app.py

@@ -0,0 +1,200 @@
+#import spaces
+import json
+import subprocess
+import os
+import sys
+
+def run_command(command):
+    process = subprocess.Popen(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True)
+    output, error = process.communicate()
+    if process.returncode != 0:
+        print(f"Error executing command: {command}")
+        print(error.decode('utf-8'))
+        exit(1)
+    return output.decode('utf-8')
+
+def install_packages():
+    # Clone the repository with submodules
+    run_command("git clone --recurse-submodules https://github.com/abetlen/llama-cpp-python.git")
+    
+    # Change to the cloned directory
+    os.chdir("llama-cpp-python")
+    
+    # Checkout the specific commit in the llama.cpp submodule
+    os.chdir("vendor/llama.cpp")
+    run_command("git checkout 50e0535")
+    os.chdir("../..")
+    
+    # Upgrade pip
+    run_command("pip install --upgrade pip")
+    
+    # Install all optional dependencies
+    run_command("pip install -e .[all]")
+    
+    # Clear the local build cache
+    run_command("make clean")
+    
+    # Reinstall the package
+    run_command("pip install -e .")
+
+    # Install llama-cpp-agent
+    run_command("pip install llama-cpp-agent")
+    
+    print("Installation complete!")
+
+try:
+    install_packages()
+    
+    # If installation is successful, import the libraries
+    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
+    
+    print("Libraries imported successfully!")
+except Exception as e:
+    print(f"Installation failed or libraries couldn't be imported: {str(e)}")
+    sys.exit(1)
+    
+import gradio as gr
+from huggingface_hub import hf_hub_download
+
+hf_hub_download(
+    repo_id="MaziyarPanahi/Mistral-Nemo-Instruct-2407-GGUF",
+    filename="Mistral-Nemo-Instruct-2407.Q5_K_M.gguf",
+    local_dir="./models"
+)
+
+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,
+):
+
+    
+    chat_template = MessagesFormatterType.MISTRAL
+
+    global llm
+    global llm_model
+    
+    if llm is None or llm_model != model:
+        llm = Llama(
+            model_path=f"models/{model}",
+            flash_attn=True,
+            n_gpu_layers=81,
+            n_batch=1024,
+            n_ctx=32768,
+        )
+        llm_model = model
+
+    provider = LlamaCppPythonProvider(llm)
+
+    agent = LlamaCppAgent(
+        provider,
+        system_prompt=f"{system_message}",
+        predefined_messages_formatter_type=chat_template,
+        debug_output=True
+    )
+    
+    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()
+
+    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 = agent.get_chat_response(
+        message,
+        llm_sampling_settings=settings,
+        chat_history=messages,
+        returns_streaming_generator=True,
+        print_output=False
+    )
+    
+    outputs = ""
+    for output in stream:
+        outputs += output
+        yield outputs
+
+description = """<p><center>
+<a href="https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407" target="_blank">[Instruct Model]</a>
+<a href="https://huggingface.co/mistralai/Mistral-Nemo-Base-2407" target="_blank">[Base Model]</a>
+<a href="https://huggingface.co/second-state/Mistral-Nemo-Instruct-2407-GGUF" target="_blank">[GGUF Version]</a>
+</center></p>
+"""
+
+demo = gr.ChatInterface(
+    respond,
+    additional_inputs=[
+        gr.Dropdown([
+                'Mistral-Nemo-Instruct-2407.Q5_K_M.gguf'
+            ],
+            value="Mistral-Nemo-Instruct-2407.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 Mistral-NeMo using llama.cpp", 
+    description=description,
+    chatbot=gr.Chatbot(
+        scale=1, 
+        likeable=False,
+        show_copy_button=True
+    )
+)
+
+if __name__ == "__main__":
+    demo.launch(debug=True)