Create app.py

app.py

@@ -0,0 +1,72 @@
+import gradio as gr
+import subprocess
+import pandas as pd
+import uuid
+import os
+
+def run_prediction(data_path, model_path, prediction_type="classification", output_path="results.csv"):
+    command = [
+        "python", model_path,
+        "--data_path", data_path,
+        "--checkpoint_path", model_path.replace("predict.py", "../models/fusion_model.pt" if "mod" in model_path else "../models/vanilla_model.pt"),
+        "--prediction_type", prediction_type,
+        "--save_dir", ".",
+        "--preds_path", output_path
+    ]
+    subprocess.run(command, check=True)
+    return pd.read_csv(output_path)
+
+def predict_from_smiles(smiles, model_choice, dataset_choice):
+    temp_id = uuid.uuid4().hex
+    temp_csv = f"data/temp_{temp_id}.csv"
+    out_csv = f"data/output_{temp_id}.csv"
+    df = pd.DataFrame([{"smiles": smiles, "compound_name": "molecule"}])
+    df.to_csv(temp_csv, index=False)
+
+    model_path = "chemprop_mod/predict.py" if model_choice == "Fusion (GNN + Transformer)" else "chemprop/predict.py"
+    
+    try:
+        predictions = run_prediction(temp_csv, model_path, output_path=out_csv)
+        return predictions
+    except subprocess.CalledProcessError as e:
+        return f"Error: {str(e)}"
+
+def predict_from_file(file_obj, model_choice, dataset_choice):
+    temp_id = uuid.uuid4().hex
+    file_path = f"data/uploaded_{temp_id}.csv"
+    out_csv = f"data/output_file_{temp_id}.csv"
+    
+    with open(file_path, "wb") as f:
+        f.write(file_obj.read())
+
+    model_path = "chemprop_mod/predict.py" if model_choice == "Fusion (GNN + Transformer)" else "chemprop/predict.py"
+
+    try:
+        predictions = run_prediction(file_path, model_path, output_path=out_csv)
+        return predictions
+    except subprocess.CalledProcessError as e:
+        return f"Error: {str(e)}"
+
+with gr.Blocks() as demo:
+    gr.Markdown("## 🧪 Drug Property Prediction with Fusion Models")
+    gr.Markdown("Choose prediction input type and compare Chemprop vs Fusion model")
+
+    with gr.Tab("Single SMILES"):
+        with gr.Row():
+            smiles_input = gr.Textbox(label="Enter SMILES string")
+            model_select = gr.Radio(["Vanilla Chemprop", "Fusion (GNN + Transformer)"], label="Model")
+            dataset_select = gr.Dropdown(["BBBP", "BACE"], label="Dataset")
+        predict_button = gr.Button("Predict")
+        result_output = gr.Dataframe(label="Prediction Result")
+        predict_button.click(fn=predict_from_smiles, inputs=[smiles_input, model_select, dataset_select], outputs=result_output)
+
+    with gr.Tab("Upload File"):
+        with gr.Row():
+            file_input = gr.File(label="Upload CSV File")
+            model_select_file = gr.Radio(["Vanilla Chemprop", "Fusion (GNN + Transformer)"], label="Model")
+            dataset_select_file = gr.Dropdown(["BBBP", "BACE"], label="Dataset")
+        predict_button_file = gr.Button("Predict")
+        result_output_file = gr.Dataframe(label="Prediction Result")
+        predict_button_file.click(fn=predict_from_file, inputs=[file_input, model_select_file, dataset_select_file], outputs=result_output_file)
+
+demo.launch()