File size: 2,459 Bytes
eb50141 7c2e9ad eb50141 4bf96b6 fa14f0d 4bf96b6 eb50141 4bf96b6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 |
import gradio as gr
from llama_cpp import Llama
from huggingface_hub import hf_hub_download
# Download the model
model_name = "Mykes/med_tinyllama_gguf"
filename = "unsloth.Q4_K_M.gguf"
model_path = hf_hub_download(repo_id=model_name, filename=filename)
# Initialize the model
# model = Llama(model_path=model_path, n_ctx=2048, n_threads=4, n_batch=32, use_mmap=True, use_mlock=True, rope_freq_base=10000, rope_freq_scale=1.0)
model = Llama(model_path=model_path, n_ctx=256, n_threads=2, n_batch=8, use_mlock=True)
# def preload_model(model, preload_tokens=1024):
# # Dummy call to load model into RAM by accessing parts of it
# try:
# dummy_input = " " * preload_tokens
# _ = model(dummy_input, max_tokens=1)
# print("Model preloaded into RAM.")
# except Exception as e:
# print(f"Error preloading model: {e}")
# # Preload the model into RAM
# preload_model(model)
def respond(
message,
history: list[tuple[str, str]],
system_message,
max_tokens,
temperature,
top_p,
):
history = history[-3:]
# Construct the prompt
prompt = f"<s>{system_message}\n\n"
for user_msg, assistant_msg in history:
prompt += f"<|user|>{user_msg}<|end|></s> <|assistant|>{assistant_msg}<|end|></s>"
prompt += f"<|user|>{message}<|end|></s> <|assistant|>"
# Generate response
response = ""
for token in model(
prompt,
max_tokens=max_tokens,
temperature=temperature,
top_p=top_p,
stream=True,
stop=["<|end|>", "</s>"]
):
response += token['choices'][0]['text']
yield response.strip()
# Create the Gradio interface
demo = gr.ChatInterface(
respond,
undo_btn="Отменить",
clear_btn="Очистить",
additional_inputs=[
# gr.Textbox(value="You are a friendly medical assistant.", label="System message"),
gr.Textbox(value="", label="System message"),
gr.Slider(minimum=128, maximum=4096, value=2048, step=1, label="Max new tokens"),
gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.9,
step=0.05,
label="Top-p (nucleus sampling)",
),
],
title="Med TinyLlama Chat",
description="Chat with the Med TinyLlama model for medical information.",
)
if __name__ == "__main__":
demo.launch() |