feat: first version

app.py

@@ -0,0 +1,49 @@
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+import torch
+
+
+model_list = [
+    "ruanchaves/mdeberta-v3-base-assin2-similarity",
+    "ruanchaves/bert-base-portuguese-cased-assin2-similarity",
+    "ruanchaves/bert-large-portuguese-cased-assin2-similarity",
+    "ruanchaves/mdeberta-v3-base-assin-similarity",
+    "ruanchaves/bert-base-portuguese-cased-assin-similarity",
+    "ruanchaves/bert-large-portuguese-cased-assin-similarity",
+]
+
+model_array = []
+
+for model_name in model_list:
+    row = {}
+    row["tokenizer"] = AutoTokenizer.from_pretrained(model_name)
+    row["model"] = AutoModelForSequenceClassification.from_pretrained(model_name)
+    model_array.append(row)
+
+
+def similarity(s1, s2):
+    scores = []
+    for row in model_array:
+        tokenizer = row["tokenizer"]
+        model = row["model"]
+        model_input = tokenizer(*([s1, s1], [s2, s1]), padding=True, return_tensors="pt")
+        with torch.no_grad():
+            output = model(**model_input)
+            score = output[0][0].item()
+            scores.append(score)
+    return sum(scores) / len(scores)
+
+
+inputs = [
+    gr.inputs.Textbox(label="Text 1"),
+    gr.inputs.Textbox(label="Text 2")
+]
+
+outputs = gr.outputs.Textbox(label="Similarity Score")
+
+
+gr.Interface(fn=similarity, inputs=inputs, outputs=outputs, title="Semantic Similarity", 
+             description="Calculates semantic similarity between two pieces of text using multiple pre-trained models.",
+             examples=[["A quem é atribuida a invenção do ábaco?", "A primeira ferramenta conhecida para a computação foi o ábaco, cuja invenção é atribuída a habitantes da Mesopotâmia, em torno de 2700–2300 a.C.."],
+                      ["I love pizza", "Pizza is my favorite food"],
+                      ["I hate cats", "I love dogs"]]).launch()