Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import gradio as gr
|
| 3 |
+
from llama_cpp import Llama
|
| 4 |
+
from huggingface_hub import hf_hub_download
|
| 5 |
+
|
| 6 |
+
model = Llama(
|
| 7 |
+
model_path=hf_hub_download(
|
| 8 |
+
repo_id=os.environ.get("REPO_ID", "bartowski/QwQ-32B-Preview-GGUF"),
|
| 9 |
+
filename=os.environ.get("MODEL_FILE", "QwQ-32B-Preview-Q3_K_L.gguf"),
|
| 10 |
+
)
|
| 11 |
+
)
|
| 12 |
+
|
| 13 |
+
DESCRIPTION = '''
|
| 14 |
+
# QwQ-32B-Preview
|
| 15 |
+
Qwen/QwQ-32B-Preview: an experimental research model developed by the Qwen Team.
|
| 16 |
+
Focused on advancing AI reasoning capabilities.
|
| 17 |
+
|
| 18 |
+
**To start a new chat**, click "clear" and start a new dialog.
|
| 19 |
+
'''
|
| 20 |
+
|
| 21 |
+
LICENSE = """
|
| 22 |
+
--- Apache 2.0 License ---
|
| 23 |
+
"""
|
| 24 |
+
|
| 25 |
+
def generate_text(message, history, max_tokens=512, temperature=0.9, top_p=0.95):
|
| 26 |
+
"""Generate a response using the Llama model."""
|
| 27 |
+
temp = ""
|
| 28 |
+
response = model.create_chat_completion(
|
| 29 |
+
messages=[{"role": "system", "content": "You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step."},
|
| 30 |
+
{"role": "user", "content": message}],
|
| 31 |
+
temperature=temperature,
|
| 32 |
+
max_tokens=max_tokens,
|
| 33 |
+
top_p=top_p,
|
| 34 |
+
stream=True,
|
| 35 |
+
)
|
| 36 |
+
for streamed in response:
|
| 37 |
+
delta = streamed["choices"][0].get("delta", {})
|
| 38 |
+
text_chunk = delta.get("content", "")
|
| 39 |
+
temp += text_chunk
|
| 40 |
+
yield temp
|
| 41 |
+
|
| 42 |
+
with gr.Blocks() as demo:
|
| 43 |
+
gr.Markdown(DESCRIPTION)
|
| 44 |
+
|
| 45 |
+
chatbot = gr.ChatInterface(
|
| 46 |
+
generate_text,
|
| 47 |
+
title="Qwen/QwQ-32B-Preview | GGUF Demo",
|
| 48 |
+
description=" settings below if needed.",
|
| 49 |
+
examples=[
|
| 50 |
+
["How many r's are in the word strawberry?"],
|
| 51 |
+
['What is the most optimal way to do Test-Time Scaling?'],
|
| 52 |
+
['Explain to me how gravity works like I am 5!'],
|
| 53 |
+
],
|
| 54 |
+
cache_examples=False,
|
| 55 |
+
fill_height=True
|
| 56 |
+
)
|
| 57 |
+
|
| 58 |
+
with gr.Accordion("Adjust Parameters", open=False):
|
| 59 |
+
gr.Slider(minimum=512, maximum=4096, value=1024, step=1, label="Max Tokens")
|
| 60 |
+
gr.Slider(minimum=0.1, maximum=1.5, value=0.9, step=0.1, label="Temperature")
|
| 61 |
+
gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
|
| 62 |
+
|
| 63 |
+
gr.Markdown(LICENSE)
|
| 64 |
+
|
| 65 |
+
if __name__ == "__main__":
|
| 66 |
+
demo.launch()
|