Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import torch
|
| 2 |
+
from transformers import AutoTokenizer, AutoModel
|
| 3 |
+
from torch.nn.functional import cosine_similarity
|
| 4 |
+
import gradio as gr
|
| 5 |
+
|
| 6 |
+
model_name = 'bert-base-multilingual-cased'
|
| 7 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 8 |
+
model = AutoModel.from_pretrained(model_name)
|
| 9 |
+
|
| 10 |
+
# Function to compute embeddings
|
| 11 |
+
def compute_embedding(text):
|
| 12 |
+
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
|
| 13 |
+
with torch.no_grad():
|
| 14 |
+
outputs = model(**inputs)
|
| 15 |
+
embedding = outputs.last_hidden_state.mean(dim=1)
|
| 16 |
+
return embedding
|
| 17 |
+
|
| 18 |
+
# Function to compute similarity between two sentences
|
| 19 |
+
def compare_sentences(text1, text2):
|
| 20 |
+
embedding1 = compute_embedding(text1)
|
| 21 |
+
embedding2 = compute_embedding(text2)
|
| 22 |
+
similarity_score = cosine_similarity(embedding1, embedding2).item()
|
| 23 |
+
return f"Similarity Score: {similarity_score:.4f}"
|
| 24 |
+
|
| 25 |
+
# Gradio interface for input
|
| 26 |
+
iface = gr.Interface(fn=compare_sentences,
|
| 27 |
+
inputs=["text", "text"],
|
| 28 |
+
outputs="text",
|
| 29 |
+
title="Sentence Similarity",
|
| 30 |
+
description="Enter two sentences to compute their similarity.")
|
| 31 |
+
|
| 32 |
+
iface.launch()
|