Spaces:
Running
Running
Initial commit
Browse files- README.md +1 -1
- app.py +53 -0
- fine-tuned_all-distilroberta-v1_quantized.onnx +3 -0
- requirements.txt +4 -0
README.md
CHANGED
|
@@ -4,7 +4,7 @@ emoji: 🐨
|
|
| 4 |
colorFrom: purple
|
| 5 |
colorTo: yellow
|
| 6 |
sdk: gradio
|
| 7 |
-
sdk_version:
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
license: mit
|
|
|
|
| 4 |
colorFrom: purple
|
| 5 |
colorTo: yellow
|
| 6 |
sdk: gradio
|
| 7 |
+
sdk_version: 3.50.0
|
| 8 |
app_file: app.py
|
| 9 |
pinned: false
|
| 10 |
license: mit
|
app.py
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import onnxruntime as ort
|
| 3 |
+
from transformers import AutoTokenizer
|
| 4 |
+
import numpy as np
|
| 5 |
+
|
| 6 |
+
# Load the ONNX model
|
| 7 |
+
onnx_model_path = "fine-tuned_all-distilroberta-v1_quantized.onnx"
|
| 8 |
+
ort_session = ort.InferenceSession(onnx_model_path)
|
| 9 |
+
|
| 10 |
+
# Load the tokenizer
|
| 11 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
| 12 |
+
"sentence-transformers/all-MiniLM-L6-v2")
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
def predict_similarity(question, candidate_answer, ai_answer):
|
| 16 |
+
# Combine question and answers
|
| 17 |
+
candidate_combined = f"Question: {question} Answer: {candidate_answer}"
|
| 18 |
+
ai_combined = f"Question: {question} Answer: {ai_answer}"
|
| 19 |
+
|
| 20 |
+
# Tokenize inputs
|
| 21 |
+
inputs = tokenizer([candidate_combined, ai_combined],
|
| 22 |
+
padding=True, truncation=True, return_tensors="np")
|
| 23 |
+
|
| 24 |
+
# Run inference
|
| 25 |
+
ort_inputs = {
|
| 26 |
+
"input_ids": inputs["input_ids"],
|
| 27 |
+
"attention_mask": inputs["attention_mask"]
|
| 28 |
+
}
|
| 29 |
+
ort_outputs = ort_session.run(None, ort_inputs)
|
| 30 |
+
|
| 31 |
+
# Calculate cosine similarity
|
| 32 |
+
embeddings = ort_outputs[0]
|
| 33 |
+
similarity = np.dot(embeddings[0], embeddings[1]) / \
|
| 34 |
+
(np.linalg.norm(embeddings[0]) * np.linalg.norm(embeddings[1]))
|
| 35 |
+
|
| 36 |
+
return float(similarity)
|
| 37 |
+
|
| 38 |
+
|
| 39 |
+
# Create Gradio interface
|
| 40 |
+
iface = gr.Interface(
|
| 41 |
+
fn=predict_similarity,
|
| 42 |
+
inputs=[
|
| 43 |
+
gr.Textbox(label="Coding Question"),
|
| 44 |
+
gr.Textbox(label="Candidate's Answer"),
|
| 45 |
+
gr.Textbox(label="AI-generated Answer")
|
| 46 |
+
],
|
| 47 |
+
outputs=gr.Number(label="Similarity Score"),
|
| 48 |
+
title="AI Code Detector",
|
| 49 |
+
description="Detect similarity between human-written and AI-generated coding answers."
|
| 50 |
+
)
|
| 51 |
+
|
| 52 |
+
# Launch the app
|
| 53 |
+
iface.launch()
|
fine-tuned_all-distilroberta-v1_quantized.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b3df341367812546bc086c5d3fa809bbde3a35dc89cc34cd5bff32dd0bd711f3
|
| 3 |
+
size 82520985
|
requirements.txt
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
gradio==3.50.0
|
| 2 |
+
onnxruntime==1.19.2
|
| 3 |
+
torch==2.4.1
|
| 4 |
+
transformers==4.44.2
|