llama-cpp-agent / app.py
pabloce's picture
Update app.py
ec06a49 verified
raw
history blame
2.15 kB
import spaces
import gradio as gr
from llama_cpp import Llama
from llama_cpp_agent import LlamaCppAgent
from llama_cpp_agent import MessagesFormatterType
from llama_cpp_agent.providers import LlamaCppPythonProvider
"""
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("cognitivecomputations/dolphin-2.8-mistral-7b-v02")
llama_model = Llama(r"Meta-Llama-3-8B.Q5_K_M.gguf", n_batch=1024, n_threads=10, n_gpu_layers=33, n_ctx=8192, verbose=False)
provider = LlamaCppPythonProvider(llama_model)
@spaces.GPU
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
):
agent = LlamaCppAgent(
provider,
system_prompt=system_message,
predefined_messages_formatter_type=MessagesFormatterType.MISTRAL,
debug_output=True
)
settings = provider.get_provider_default_settings()
settings.max_tokens = max_tokens
settings.temperature = temperature
settings.top_p = top_p
agent_output = agent.get_chat_response(message, llm_sampling_settings=settings)
yield agent_output.strip()
"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
respond,
additional_inputs=[
gr.Textbox(value="You are a helpful assistant.", 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)",
),
],
theme=gr.themes.Soft(primary_hue="green", secondary_hue="indigo", neutral_hue="zinc",font=[gr.themes.GoogleFont("Exo 2"), "ui-sans-serif", "system-ui", "sans-serif"]).set(
block_background_fill_dark="*neutral_800"
)
)
if __name__ == "__main__":
demo.launch()