Create app.py

app.py

@@ -0,0 +1,53 @@
+import gradio as gr
+import pickle
+import tensorflow as tf
+from tensorflow.keras.preprocessing.sequence import pad_sequences
+import numpy as np
+
+
+import keras
+
+
+# Load tokenizer and model
+tokenizer_path = "tokenizer.pkl"
+#model_path = r"models/toxic_classifier.keras"
+
+with open(tokenizer_path, "rb") as f:
+    tokenizer = pickle.load(f)
+model = keras.saving.load_model("hf://NightPrince/Toxic-Predict")
+
+# Label map
+label_map = {
+    0: "Child Sexual Exploitation",
+    1: "Elections",
+    2: "Non-Violent Crimes",
+    3: "Safe",
+    4: "Sex-Related Crimes",
+    5: "Suicide & Self-Harm",
+    6: "Unknown S-Type",
+    7: "Violent Crimes",
+    8: "Unsafe"
+}
+
+def classify_toxic(query, image_desc):
+    max_len = 150
+    text = query + " " + image_desc
+    seq = tokenizer.texts_to_sequences([text])
+    pad = pad_sequences(seq, maxlen=max_len, padding='post', truncating='post')
+    pred = model.predict(pad)
+    pred_label = np.argmax(pred, axis=1)[0]
+    return label_map.get(pred_label, "Unknown")
+
+iface = gr.Interface(
+    fn=classify_toxic,
+    inputs=[
+        gr.Textbox(label="Query"),
+        gr.Textbox(label="Image Description")
+    ],
+    outputs=gr.Textbox(label="Predicted Toxic Category"),
+    title="Toxic Category Classifier",
+    description="Enter a query and image description to classify the prompt into one of the toxic categories"
+)
+
+if __name__ == "__main__":
+    iface.launch()