add app

app.py

@@ -0,0 +1,46 @@
+from statistics import mean
+import random
+
+import torch
+from transformers import BertModel, BertTokenizerFast
+import numpy as np
+import torch.nn.functional as F
+import gradio as gr
+
+tokenizer = BertTokenizerFast.from_pretrained("setu4993/LaBSE")
+model = BertModel.from_pretrained("setu4993/LaBSE")
+model = model.eval()
+
+def embed(text, tokenizer, model):
+    inputs = tokenizer(text, return_tensors="pt", padding=True)
+    with torch.no_grad():
+        outputs = model(**inputs)
+    return outputs.pooler_output
+
+def similarity(embeddings_1, embeddings_2):
+    normalized_embeddings_1 = F.normalize(embeddings_1, p=2)
+    normalized_embeddings_2 = F.normalize(embeddings_2, p=2)
+    return torch.matmul(
+        normalized_embeddings_1, normalized_embeddings_2.transpose(0, 1)
+    )
+
+def semantic_sim(sentence1, sentence2):
+    em1 = embed(sentence1, tokenizer, model)
+    em2 = embed(sentence2, tokenizer, model)
+    sim = int(float(similarity(em1, em2)*5))
+    out = ""
+    if sim == 5:
+        out = "Equivalent"
+    elif sim == 4:
+        out = "Mostly equivalent, unimportant details differ"
+    elif sim == 3:
+        out = "Roughly equivalent, important details differ or are missing"
+    elif sim == 2:
+        out = "Not equivalent, but share some details"
+    elif sim == 1:
+        out = "Same general topic, but not equivalent"
+    elif sim == 0:
+        out = "Completely dissimilar"
+    return out
+
+iface = gr.Interface(fn=semantic_sim, inputs=["text", "text"], outputs=["text"]).launch()

requirements.txt

@@ -0,0 +1,5 @@
+torch
+transformers
+numpy
+gradio
+