Upload folder using huggingface_hub

app.py

@@ -629,7 +629,7 @@ def load_model_and_tokenizers():
 
 
 # --- Inference Function for Gradio ---
-def predict_iupac(smiles_string, beam_width_str, n_best_str, length_penalty_str):
+def predict_iupac(smiles_string, beam_width_str, n_best_str):
     """
     Performs SMILES to IUPAC translation using the loaded model and beam search.
     Takes string inputs from Gradio sliders/inputs and converts them.
@@ -660,26 +660,16 @@ def predict_iupac(smiles_string, beam_width_str, n_best_str, length_penalty_str)
     try:
         beam_width = int(beam_width_str)
         n_best = int(n_best_str)
-        length_penalty = float(length_penalty_str)
         if beam_width < 1 or n_best < 1 or n_best > beam_width:
             raise ValueError(
                 "Beam width and n_best must be >= 1, and n_best <= beam width."
             )
-        if length_penalty < 0:
-            logging.warning(
-                f"Length penalty {length_penalty} is negative, using 0.0 instead."
-            )
-            length_penalty = 0.0
     except ValueError as e:
         error_msg = f"Error: Invalid input parameter ({e}). Please check beam width, n_best, and length penalty values."
         logging.error(error_msg)
         # Cannot determine n_best if its input was invalid, default to 1 error line
         return f"1. {error_msg}"
 
-    logging.info(
-        f"Translating SMILES: '{smiles_input}' (Beam={beam_width}, N={n_best}, Penalty={length_penalty:.2f})"
-    )
-
     try:
         # --- Call the core translation logic ---
         # Retrieve necessary IDs from the loaded config
@@ -700,7 +690,7 @@ def predict_iupac(smiles_string, beam_width_str, n_best_str, length_penalty_str)
             pad_idx=pad_idx,
             beam_width=beam_width,
             n_best=n_best,
-            length_penalty=length_penalty,
+            length_penalty=0.0,
         )
         logging.info(f"Predictions returned: {predicted_names}")
 
@@ -801,14 +791,7 @@ n_best_input = gr.Slider(
     label="Number of Results (n_best)",
     info="How many top sequences to return (must be <= Beam Width).",
 )
-length_penalty_input = gr.Slider(
-    minimum=0.0,
-    maximum=2.0,
-    value=0.6,
-    step=0.1,
-    label="Length Penalty (alpha)",
-    info="Controls preference for sequence length. >1 favors longer, <1 favors shorter, 0 no penalty.",
-)
+
 output_text = gr.Textbox(
     label="Predicted IUPAC Name(s)", lines=5, show_copy_button=True
 )
@@ -820,7 +803,6 @@ iface = gr.Interface(
         smiles_input,
         beam_width_input,
         n_best_input,
-        length_penalty_input,
     ],
     outputs=output_text,  # Output component
     title=title,
@@ -839,9 +821,4 @@ iface = gr.Interface(
 
 # --- Launch the App ---
 if __name__ == "__main__":
-    # Launch the Gradio interface
-    # share=True generates a public link (useful for testing outside HF Spaces, but temporary)
-    # Set share=False or remove for deployment on Spaces.
-    # Use server_name="0.0.0.0" to make it accessible on the network if running locally
-    # Use auth=("username", "password") for basic authentication
-    iface.launch()  # share=True is deprecated, use launch()
+    iface.launch()