Create app.py

app.py

@@ -0,0 +1,35 @@
+import gradio as gr
+import torch
+from transformers import BertTokenizer, BertModel
+
+# Load BERT (uncased) tokenizer & model
+tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
+model = BertModel.from_pretrained("bert-base-uncased")
+
+def preprocess(text: str):
+    # remove non‑ASCII characters and lowercase
+    cleaned = text.encode("ascii", "ignore").decode().lower()
+    # tokenize
+    inputs = tokenizer(cleaned, return_tensors="pt")
+    tokens = tokenizer.convert_ids_to_tokens(inputs["input_ids"][0])
+    # get embeddings
+    with torch.no_grad():
+        outputs = model(**inputs)
+    # last_hidden_state: [1, seq_len, hidden_size] → drop batch dim
+    embeddings = outputs.last_hidden_state[0].tolist()
+    return tokens, embeddings
+
+iface = gr.Interface(
+    fn=preprocess,
+    inputs=gr.Textbox(lines=3, placeholder="Type something…"),
+    outputs=[
+        gr.Dataframe(headers=["token"], label="Tokens"),
+        gr.Dataframe(label="Embeddings")
+    ],
+    title="BERT Tokenizer + Embeddings",
+    description="Cleans input, lowercases it, then shows BERT tokens & their hidden‑state vectors."
+)
+
+if __name__ == "__main__":
+    iface.launch()
+