Add information text

app.py

@@ -126,9 +126,24 @@ def predict_toxicity(text, model, tokenizer, device, model_name):
     return prediction, inference_time
 
 def main():
-    st.set_page_config(page_title="Multi-Model Toxicity Detector", layout="wide")
-    st.title("Multi-Model Toxicity Detector")
+    st.set_page_config(page_title="Toxicity Detector Model Comparison", layout="wide")
+    st.title("Toxicity Detector Model Comparison")
 
+    # Explanation text
+    st.markdown("""
+    ### How It Works
+    This application compares various toxicity detection models to classify whether a given text is toxic or not. The models being compared include:
+    
+    - **Tiny-Toxic-Detector**: A new 2M parameter model being released soon.
+    - [**Toxic-BERT**](https://huggingface.co/unitary/toxic-bert): A 109M parameter BERT-based model.
+    - [**RoBERTa-Toxicity-Classifier**](s-nlp/roberta_toxicity_classifier): A 124M parameter RoBERTa-based model.
+    - [**Toxic-Comment-Model**](https://huggingface.co/martin-ha/toxic-comment-model): A 67M parameter DistilBERT-based model.
+    - [**ToxicChat-T5**](https://huggingface.co/lmsys/toxicchat-t5-large-v1.0): A 738M parameter T5-based model.
+
+    Simply enter the text you want to classify, and the app will provide the predictions from each model, along with the inference time.
+    Please note these models are (mostly) English-only.
+    """)
+    
     # Load models
     hf_token = os.getenv('AT')
     models, tokenizers, device = load_models_and_tokenizers(hf_token)