Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# app.py
|
| 2 |
+
# =============
|
| 3 |
+
# This is a complete app.py file for a text generation app using the Qwen/Qwen2.5-Coder-0.5B-Instruct-GGUF model.
|
| 4 |
+
# The app is built using Gradio and runs on a CPU without video memory.
|
| 5 |
+
|
| 6 |
+
# Imports
|
| 7 |
+
# =======
|
| 8 |
+
import gradio as gr
|
| 9 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
| 10 |
+
import torch
|
| 11 |
+
|
| 12 |
+
# Constants
|
| 13 |
+
# =========
|
| 14 |
+
MODEL_NAME = "Qwen/Qwen2.5-Coder-0.5B-Instruct-GGUF"
|
| 15 |
+
DEVICE = "cpu" # Ensure the model runs on CPU
|
| 16 |
+
|
| 17 |
+
# Load Model and Tokenizer
|
| 18 |
+
# ========================
|
| 19 |
+
def load_model_and_tokenizer():
|
| 20 |
+
"""
|
| 21 |
+
Load the model and tokenizer from Hugging Face.
|
| 22 |
+
"""
|
| 23 |
+
tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
|
| 24 |
+
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.bfloat16, device_map=DEVICE)
|
| 25 |
+
return tokenizer, model
|
| 26 |
+
|
| 27 |
+
tokenizer, model = load_model_and_tokenizer()
|
| 28 |
+
|
| 29 |
+
# Generate Text
|
| 30 |
+
# =============
|
| 31 |
+
def generate_text(prompt, max_length=100):
|
| 32 |
+
"""
|
| 33 |
+
Generate text based on the given prompt.
|
| 34 |
+
|
| 35 |
+
Args:
|
| 36 |
+
prompt (str): The input prompt for text generation.
|
| 37 |
+
max_length (int): The maximum length of the generated text.
|
| 38 |
+
|
| 39 |
+
Returns:
|
| 40 |
+
str: The generated text.
|
| 41 |
+
"""
|
| 42 |
+
inputs = tokenizer(prompt, return_tensors="pt").to(DEVICE)
|
| 43 |
+
outputs = model.generate(inputs.input_ids, max_length=max_length, num_return_sequences=1)
|
| 44 |
+
generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 45 |
+
return generated_text
|
| 46 |
+
|
| 47 |
+
# Gradio Interface
|
| 48 |
+
# =================
|
| 49 |
+
def gradio_interface():
|
| 50 |
+
"""
|
| 51 |
+
Create and launch the Gradio interface.
|
| 52 |
+
"""
|
| 53 |
+
iface = gr.Interface(
|
| 54 |
+
fn=generate_text,
|
| 55 |
+
inputs=[
|
| 56 |
+
gr.inputs.Textbox(lines=2, placeholder="Enter your prompt here..."),
|
| 57 |
+
gr.inputs.Slider(minimum=50, maximum=500, step=10, default=100, label="Max Length")
|
| 58 |
+
],
|
| 59 |
+
outputs="text",
|
| 60 |
+
title="Qwen2.5-Coder-0.5B-Instruct-GGUF Text Generation",
|
| 61 |
+
description="Generate text using the Qwen2.5-Coder-0.5B-Instruct-GGUF model."
|
| 62 |
+
)
|
| 63 |
+
iface.launch()
|
| 64 |
+
|
| 65 |
+
# Main
|
| 66 |
+
# ====
|
| 67 |
+
if __name__ == "__main__":
|
| 68 |
+
gradio_interface()
|
| 69 |
+
|
| 70 |
+
# Dependencies
|
| 71 |
+
# =============
|
| 72 |
+
# The following dependencies are required to run this app:
|
| 73 |
+
# - transformers
|
| 74 |
+
# - gradio
|
| 75 |
+
# - torch
|
| 76 |
+
#
|
| 77 |
+
# You can install these dependencies using pip:
|
| 78 |
+
# pip install transformers gradio torch
|