File size: 2,352 Bytes
4667b5a 769b533 60faee4 6d0db7f 1d873a4 07c2cc6 1d873a4 3a75081 1d873a4 5fb8783 1d873a4 5fb8783 1d873a4 07c2cc6 1d873a4 6d0db7f 1d873a4 6d0db7f 60faee4 1d873a4 60faee4 8314b3e 60faee4 1d873a4 60faee4 1d873a4 60faee4 1d873a4 |
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 |
import spaces
import threading
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
# Load the model and tokenizer locally
model_name = "kz919/QwQ-0.5B-Distilled-SFT"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda")
# Define the function to handle chat responses
@spaces.GPU
def respond(message, history: list[tuple[str, str]], system_message, max_tokens, temperature, top_p):
# Prepare the prompt by combining history and system messages
msg = [
{"role": "system", "content": system_message}
]
for user_input, assistant_response in history:
msg.extend(
{"role": "user", "content": user_input},
{"role": "assistant", "content": assistant_response}
)
msg.append({"role": "user", "content": message})
prompt = tokenizer.apply_chat_template(
msg,
tokenize=False,
add_generation_prompt=True
)
# Tokenize the input prompt
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
# Use a thread to run the generation in parallel
generation_thread = threading.Thread(
target=model.generate,
kwargs=dict(
inputs=inputs.input_ids,
max_length=max_tokens,
streamer=streamer,
do_sample=True,
temperature=temperature,
top_p=top_p,
pad_token_id=tokenizer.eos_token_id,
),
)
generation_thread.start()
# Stream the tokens as they are generated
for new_text in streamer:
yield new_text
# Create the Gradio interface
demo = gr.ChatInterface(
respond,
additional_inputs=[
gr.Textbox(value="You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step.", 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)"),
],
)
# Launch the Gradio app
if __name__ == "__main__":
demo.launch()
|