Spaces:
Runtime error
Runtime error
File size: 3,719 Bytes
35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f 35af015 98bdd9f |
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 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 |
import gradio as gr
from transformers import AutoTokenizer, AutoModel
# from MMD_calculate import mmd_two_sample_baseline # Adjust path based on your structure
# from utils_MMD import extract_features # Example helper from your utils
MINIMUM_TOKENS = 64
def count_tokens(text, tokenizer):
return len(tokenizer(text).input_ids)
def run_test_power(model_name, tokenizer_name, real_text, generated_text, N):
"""
Runs the test power calculation for provided real and generated texts.
"""
# load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_name).cuda()
model = AutoModel.from_pretrained(model)
if count_tokens(real_text, tokenizer) < MINIMUM_TOKENS or count_tokens(generated_text, tokenizer) < MINIMUM_TOKENS:
return "Too short length. Need minimum 64 tokens to calculated Test Power."
# Extract features
fea_real_ls = extract_features(model_name, tokenizer_name, [real_text])
fea_generated_ls = extract_features(model_name, tokenizer_name, [generated_text])
# Calculate test power list
test_power_ls = mmd_two_sample_baseline(fea_real_ls, fea_generated_ls, N=10)
# Compute the average test power value
power_test_value = sum(test_power_ls) / len(test_power_ls)
# Classify the text
if power_test_value < threshold:
return "Prediction: Human"
else:
return "Prediction: AI"
css = """
#header { text-align: center; font-size: 1.5em; margin-bottom: 20px; }
#output-text { font-weight: bold; font-size: 1.2em; }
"""
# Gradio App
with gr.Blocks(css=css) as app:
with gr.Row():
gr.HTML('<div id="header">Human or AI Text Detector</div>')
with gr.Row():
gr.Markdown(
"""
[Paper](https://openreview.net/forum?id=z9j7wctoGV) | [Code](https://github.com/xLearn-AU/R-Detect) | [Contact](mailto:[email protected])
"""
)
with gr.Row():
input_text = gr.Textbox(
label="Input Text",
placeholder="Enter the text to check",
lines=8,
)
with gr.Row():
model_name = gr.Dropdown(
["gpt2-medium", "gpt2-large", "t5-large", "t5-small", "roberta-base", "roberta-base-openai-detector", "falcon-rw-1b"],
label="Select Model",
value="gpt2-medium",
)
with gr.Row():
submit_button = gr.Button("Run Detection", variant="primary")
clear_button = gr.Button("Clear", variant="secondary")
with gr.Row():
output = gr.Textbox(
label = "Prediction",
placeholder = "Prediction: Human or AI",
elem_id = "output-text",
)
with gr.Accordion("Disclaimer", open=False):
gr.Markdown(
"""
- **Disclaimer**: This tool is for demonstration purposes only. It is not a foolproof AI detector.
- **Accuracy**: Results may vary based on input length and quality.
"""
)
with gr.Accordion("Citations", open=False):
gr.Markdown(
"""
```
@inproceedings{zhangs2024MMDMP,
title={Detecting Machine-Generated Texts by Multi-Population Aware Optimization for Maximum Mean Discrepancy},
author={Zhang, Shuhai and Song, Yiliao and Yang, Jiahao and Li, Yuanqing and Han, Bo and Tan, Mingkui},
booktitle = {International Conference on Learning Representations (ICLR)},
year={2024}
}
```
"""
)
submit_button.click(detect_text, inputs=[input_text, model_name], outputs=output)
clear_button.click(lambda: ("", ""), inputs=[], outputs=[input_text, output])
app.launch()
|