Spaces:
				
			
			
	
			
			
		Running
		
			on 
			
			Zero
	
	
	
			
			
	
	
	
	
		
		
		Running
		
			on 
			
			Zero
	Merge branch 'citation'
Browse files
    	
        app.py
    CHANGED
    
    | 
         @@ -28,13 +28,15 @@ except ImportError as e: 
     | 
|
| 28 | 
         
             
                raise ImportError(f"Failed to import InstaNovo components: {e}")
         
     | 
| 29 | 
         | 
| 30 | 
         
             
            # --- Configuration ---
         
     | 
| 31 | 
         
            -
            MODEL_ID = "instanovo-v1.1.0" 
     | 
| 32 | 
         
             
            KNAPSACK_DIR = Path("./knapsack_cache")
         
     | 
| 33 | 
         
            -
            DEFAULT_CONFIG_PATH = Path( 
     | 
| 
         | 
|
| 
         | 
|
| 34 | 
         | 
| 35 | 
         
             
            # Determine device
         
     | 
| 36 | 
         
             
            DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
         
     | 
| 37 | 
         
            -
            FP16 = DEVICE == "cuda" 
     | 
| 38 | 
         | 
| 39 | 
         
             
            # --- Global Variables (Load Model and Knapsack Once) ---
         
     | 
| 40 | 
         
             
            MODEL: InstaNovo | None = None
         
     | 
| 
         @@ -78,9 +80,9 @@ def load_model_and_knapsack(): 
     | 
|
| 78 | 
         | 
| 79 | 
         
             
                # --- Knapsack Handling ---
         
     | 
| 80 | 
         
             
                knapsack_exists = (
         
     | 
| 81 | 
         
            -
                    (KNAPSACK_DIR / "parameters.pkl").exists() 
     | 
| 82 | 
         
            -
                    (KNAPSACK_DIR / "masses.npy").exists() 
     | 
| 83 | 
         
            -
                    (KNAPSACK_DIR / "chart.npy").exists()
         
     | 
| 84 | 
         
             
                )
         
     | 
| 85 | 
         | 
| 86 | 
         
             
                if knapsack_exists:
         
     | 
| 
         @@ -96,11 +98,15 @@ def load_model_and_knapsack(): 
     | 
|
| 96 | 
         
             
                if not knapsack_exists:
         
     | 
| 97 | 
         
             
                    logger.info("Knapsack not found or failed to load. Generating knapsack...")
         
     | 
| 98 | 
         
             
                    if RESIDUE_SET is None:
         
     | 
| 99 | 
         
            -
             
     | 
| 
         | 
|
| 
         | 
|
| 100 | 
         
             
                    try:
         
     | 
| 101 | 
         
             
                        # Prepare residue masses for knapsack generation (handle negative/zero masses)
         
     | 
| 102 | 
         
             
                        residue_masses_knapsack = dict(RESIDUE_SET.residue_masses.copy())
         
     | 
| 103 | 
         
            -
                        negative_residues = [ 
     | 
| 
         | 
|
| 
         | 
|
| 104 | 
         
             
                        if negative_residues:
         
     | 
| 105 | 
         
             
                            logger.info(f"Warning: Non-positive masses found in residues: {negative_residues}. "
         
     | 
| 106 | 
         
             
                                  "Excluding from knapsack generation.")
         
     | 
| 
         @@ -108,19 +114,19 @@ def load_model_and_knapsack(): 
     | 
|
| 108 | 
         
             
                                del residue_masses_knapsack[res]
         
     | 
| 109 | 
         
             
                        # Remove special tokens explicitly if they somehow got mass
         
     | 
| 110 | 
         
             
                        for special_token in RESIDUE_SET.special_tokens:
         
     | 
| 111 | 
         
            -
             
     | 
| 112 | 
         
            -
             
     | 
| 113 | 
         | 
| 114 | 
         
             
                        # Ensure residue indices used match those without special/negative masses
         
     | 
| 115 | 
         
             
                        valid_residue_indices = {
         
     | 
| 116 | 
         
            -
                            res: idx 
     | 
| 
         | 
|
| 117 | 
         
             
                            if res in residue_masses_knapsack
         
     | 
| 118 | 
         
             
                        }
         
     | 
| 119 | 
         | 
| 120 | 
         
            -
             
     | 
| 121 | 
         
             
                        KNAPSACK = Knapsack.construct_knapsack(
         
     | 
| 122 | 
         
             
                            residue_masses=residue_masses_knapsack,
         
     | 
| 123 | 
         
            -
                            residue_indices=valid_residue_indices, 
     | 
| 124 | 
         
             
                            max_mass=MAX_MASS,
         
     | 
| 125 | 
         
             
                            mass_scale=MASS_SCALE,
         
     | 
| 126 | 
         
             
                        )
         
     | 
| 
         @@ -135,6 +141,7 @@ def load_model_and_knapsack(): 
     | 
|
| 135 | 
         
             
            # Load the model and knapsack when the script starts
         
     | 
| 136 | 
         
             
            load_model_and_knapsack()
         
     | 
| 137 | 
         | 
| 
         | 
|
| 138 | 
         
             
            def create_inference_config(
         
     | 
| 139 | 
         
             
                input_path: str,
         
     | 
| 140 | 
         
             
                output_path: str,
         
     | 
| 
         @@ -143,7 +150,7 @@ def create_inference_config( 
     | 
|
| 143 | 
         
             
                """Creates the OmegaConf DictConfig needed for prediction."""
         
     | 
| 144 | 
         
             
                # Load default config if available, otherwise create from scratch
         
     | 
| 145 | 
         
             
                if DEFAULT_CONFIG_PATH.exists():
         
     | 
| 146 | 
         
            -
             
     | 
| 147 | 
         
             
                else:
         
     | 
| 148 | 
         
             
                     logger.info(f"Warning: Default config not found at {DEFAULT_CONFIG_PATH}. Using minimal config.")
         
     | 
| 149 | 
         
             
                     # Create a minimal config if default is missing
         
     | 
| 
         @@ -206,7 +213,9 @@ def create_inference_config( 
     | 
|
| 206 | 
         
             
                    cfg_overrides["use_knapsack"] = False
         
     | 
| 207 | 
         
             
                elif "Knapsack" in decoding_method:
         
     | 
| 208 | 
         
             
                    if KNAPSACK is None:
         
     | 
| 209 | 
         
            -
                        raise gr.Error( 
     | 
| 
         | 
|
| 
         | 
|
| 210 | 
         
             
                    cfg_overrides["num_beams"] = 5
         
     | 
| 211 | 
         
             
                    cfg_overrides["use_knapsack"] = True
         
     | 
| 212 | 
         
             
                    cfg_overrides["knapsack_path"] = str(KNAPSACK_DIR)
         
     | 
| 
         @@ -223,9 +232,9 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 223 | 
         
             
                Main function to load data, run prediction, and return results.
         
     | 
| 224 | 
         
             
                """
         
     | 
| 225 | 
         
             
                if MODEL is None or RESIDUE_SET is None or MODEL_CONFIG is None:
         
     | 
| 226 | 
         
            -
             
     | 
| 227 | 
         
            -
             
     | 
| 228 | 
         
            -
             
     | 
| 229 | 
         | 
| 230 | 
         
             
                if input_file is None:
         
     | 
| 231 | 
         
             
                    raise gr.Error("Please upload a mass spectrometry file.")
         
     | 
| 
         @@ -248,17 +257,18 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 248 | 
         
             
                    try:
         
     | 
| 249 | 
         
             
                        sdf = SpectrumDataFrame.load(
         
     | 
| 250 | 
         
             
                            config.data_path,
         
     | 
| 251 | 
         
            -
                            lazy=False, 
     | 
| 252 | 
         
            -
                            is_annotated=False, 
     | 
| 253 | 
         
             
                            column_mapping=config.get("column_map", None),
         
     | 
| 254 | 
         
             
                            shuffle=False,
         
     | 
| 255 | 
         
            -
                            verbose=True 
     | 
| 256 | 
         
             
                        )
         
     | 
| 257 | 
         
             
                        # Apply charge filter like in CLI
         
     | 
| 258 | 
         
             
                        original_size = len(sdf)
         
     | 
| 259 | 
         
             
                        max_charge = config.get("max_charge", 10)
         
     | 
| 260 | 
         
             
                        sdf.filter_rows(
         
     | 
| 261 | 
         
            -
                            lambda row: (row["precursor_charge"] <= max_charge) 
     | 
| 
         | 
|
| 262 | 
         
             
                        )
         
     | 
| 263 | 
         
             
                        if len(sdf) < original_size:
         
     | 
| 264 | 
         
             
                            logger.info(f"Warning: Filtered {original_size - len(sdf)} spectra with charge > {max_charge} or <= 0.")
         
     | 
| 
         @@ -275,16 +285,17 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 275 | 
         
             
                        sdf,
         
     | 
| 276 | 
         
             
                        RESIDUE_SET,
         
     | 
| 277 | 
         
             
                        MODEL_CONFIG.get("n_peaks", 200),
         
     | 
| 278 | 
         
            -
                        return_str=True, 
     | 
| 279 | 
         
             
                        annotated=False,
         
     | 
| 280 | 
         
            -
                        pad_spectrum_max_length=config.get("compile_model", False) 
     | 
| 
         | 
|
| 281 | 
         
             
                        bin_spectra=config.get("conv_peak_encoder", False),
         
     | 
| 282 | 
         
             
                    )
         
     | 
| 283 | 
         
             
                    dl = DataLoader(
         
     | 
| 284 | 
         
             
                        ds,
         
     | 
| 285 | 
         
             
                        batch_size=config.batch_size,
         
     | 
| 286 | 
         
            -
                        num_workers=0, 
     | 
| 287 | 
         
            -
                        shuffle=False, 
     | 
| 288 | 
         
             
                        collate_fn=collate_batch,
         
     | 
| 289 | 
         
             
                    )
         
     | 
| 290 | 
         | 
| 
         @@ -293,8 +304,10 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 293 | 
         
             
                    decoder: Decoder
         
     | 
| 294 | 
         
             
                    if config.use_knapsack:
         
     | 
| 295 | 
         
             
                        if KNAPSACK is None:
         
     | 
| 296 | 
         
            -
             
     | 
| 297 | 
         
            -
             
     | 
| 
         | 
|
| 
         | 
|
| 298 | 
         
             
                        # KnapsackBeamSearchDecoder doesn't directly load from path in this version?
         
     | 
| 299 | 
         
             
                        # We load Knapsack globally, so just pass it.
         
     | 
| 300 | 
         
             
                        # If it needed path: decoder = KnapsackBeamSearchDecoder.from_file(model=MODEL, path=config.knapsack_path)
         
     | 
| 
         @@ -316,15 +329,22 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 316 | 
         
             
                    # 5. Run Prediction Loop (Adapted from instanovo/transformer/predict.py)
         
     | 
| 317 | 
         
             
                    logger.info("Starting prediction...")
         
     | 
| 318 | 
         
             
                    start_time = time.time()
         
     | 
| 319 | 
         
            -
                    results_list: list[ 
     | 
| 
         | 
|
| 
         | 
|
| 320 | 
         | 
| 321 | 
         
             
                    for i, batch in enumerate(dl):
         
     | 
| 322 | 
         
            -
                        spectra, precursors, spectra_mask, _, _ =  
     | 
| 
         | 
|
| 
         | 
|
| 323 | 
         
             
                        spectra = spectra.to(DEVICE)
         
     | 
| 324 | 
         
             
                        precursors = precursors.to(DEVICE)
         
     | 
| 325 | 
         
             
                        spectra_mask = spectra_mask.to(DEVICE)
         
     | 
| 326 | 
         | 
| 327 | 
         
            -
                        with  
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 328 | 
         
             
                            # Beam search decoder might return list[list[ScoredSequence]] if return_beam=True
         
     | 
| 329 | 
         
             
                            # Greedy decoder returns list[ScoredSequence]
         
     | 
| 330 | 
         
             
                            # KnapsackBeamSearchDecoder returns list[ScoredSequence] or list[list[ScoredSequence]]
         
     | 
| 
         @@ -334,9 +354,12 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 334 | 
         
             
                                beam_size=config.num_beams,
         
     | 
| 335 | 
         
             
                                max_length=config.max_length,
         
     | 
| 336 | 
         
             
                                # Knapsack/Beam Search specific params if needed
         
     | 
| 337 | 
         
            -
                                mass_tolerance=config.get("filter_precursor_ppm", 20) 
     | 
| 338 | 
         
            -
                                 
     | 
| 339 | 
         
            -
                                 
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 340 | 
         
             
                            )
         
     | 
| 341 | 
         
             
                        results_list.extend(batch_predictions) # Should be list[ScoredSequence] or list[list]
         
     | 
| 342 | 
         
             
                        logger.info(f"Processed batch {i+1}/{len(dl)}")
         
     | 
| 
         @@ -349,26 +372,30 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 349 | 
         
             
                    output_data = []
         
     | 
| 350 | 
         
             
                    # Use sdf index columns + prediction results
         
     | 
| 351 | 
         
             
                    index_cols = [col for col in config.index_columns if col in sdf.df.columns]
         
     | 
| 352 | 
         
            -
                    base_df_pd = sdf.df.select(index_cols).to_pandas() 
     | 
| 353 | 
         | 
| 354 | 
         
             
                    metrics_calc = Metrics(RESIDUE_SET, config.isotope_error_range)
         
     | 
| 355 | 
         | 
| 356 | 
         
             
                    for i, res in enumerate(results_list):
         
     | 
| 357 | 
         
            -
                        row_data = base_df_pd.iloc[i].to_dict() 
     | 
| 358 | 
         
             
                        if isinstance(res, ScoredSequence) and res.sequence:
         
     | 
| 359 | 
         
             
                            sequence_str = "".join(res.sequence)
         
     | 
| 360 | 
         
             
                            row_data["prediction"] = sequence_str
         
     | 
| 361 | 
         
             
                            row_data["log_probability"] = f"{res.sequence_log_probability:.4f}"
         
     | 
| 362 | 
         
             
                            # Use metrics to calculate delta mass ppm for the top prediction
         
     | 
| 363 | 
         
             
                            try:
         
     | 
| 364 | 
         
            -
             
     | 
| 365 | 
         
            -
             
     | 
| 366 | 
         
            -
             
     | 
| 367 | 
         
            -
             
     | 
| 368 | 
         
            -
             
     | 
| 369 | 
         
            -
             
     | 
| 370 | 
         
            -
             
     | 
| 371 | 
         
            -
             
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 372 | 
         
             
                            except Exception as e:
         
     | 
| 373 | 
         
             
                                 logger.info(f"Warning: Could not calculate delta mass for prediction {i}: {e}")
         
     | 
| 374 | 
         
             
                                 row_data["delta_mass_ppm"] = "N/A"
         
     | 
| 
         @@ -382,7 +409,14 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 382 | 
         
             
                    output_df = pl.DataFrame(output_data)
         
     | 
| 383 | 
         | 
| 384 | 
         
             
                    # Ensure specific columns are present and ordered
         
     | 
| 385 | 
         
            -
                    display_cols = [ 
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 386 | 
         
             
                    final_display_cols = []
         
     | 
| 387 | 
         
             
                    for col in display_cols:
         
     | 
| 388 | 
         
             
                        if col in output_df.columns:
         
     | 
| 
         @@ -397,7 +431,6 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 397 | 
         | 
| 398 | 
         
             
                    output_df_display = output_df.select(final_display_cols)
         
     | 
| 399 | 
         | 
| 400 | 
         
            -
             
     | 
| 401 | 
         
             
                    # 7. Save full results to CSV
         
     | 
| 402 | 
         
             
                    logger.info(f"Saving results to {output_csv_path}...")
         
     | 
| 403 | 
         
             
                    output_df.write_csv(output_csv_path)
         
     | 
| 
         @@ -413,6 +446,7 @@ def predict_peptides(input_file, decoding_method): 
     | 
|
| 413 | 
         
             
                    # Re-raise as Gradio error
         
     | 
| 414 | 
         
             
                    raise gr.Error(f"Prediction failed: {e}")
         
     | 
| 415 | 
         | 
| 
         | 
|
| 416 | 
         
             
            # --- Gradio Interface ---
         
     | 
| 417 | 
         
             
            css = """
         
     | 
| 418 | 
         
             
            .gradio-container { font-family: sans-serif; }
         
     | 
| 
         @@ -422,7 +456,9 @@ footer { display: none !important; } 
     | 
|
| 422 | 
         
             
            .logo-container img { margin-bottom: 1rem; }
         
     | 
| 423 | 
         
             
            """
         
     | 
| 424 | 
         | 
| 425 | 
         
            -
            with gr.Blocks( 
     | 
| 
         | 
|
| 
         | 
|
| 426 | 
         
             
                # --- Logo Display ---
         
     | 
| 427 | 
         
             
                gr.Markdown(
         
     | 
| 428 | 
         
             
                    """
         
     | 
| 
         @@ -430,7 +466,7 @@ with gr.Blocks(css=css, theme=gr.themes.Default(primary_hue="blue", secondary_hu 
     | 
|
| 430 | 
         
             
                      <img src='/gradio_api/file=assets/instanovo.svg' alt="InstaNovo Logo" width="300" style="display: block; margin: 0 auto;">
         
     | 
| 431 | 
         
             
                    </div>
         
     | 
| 432 | 
         
             
                    """,
         
     | 
| 433 | 
         
            -
                    elem_classes="logo-container" 
     | 
| 434 | 
         
             
                )
         
     | 
| 435 | 
         | 
| 436 | 
         
             
                # --- App Content ---
         
     | 
| 
         @@ -445,38 +481,57 @@ with gr.Blocks(css=css, theme=gr.themes.Default(primary_hue="blue", secondary_hu 
     | 
|
| 445 | 
         
             
                    with gr.Column(scale=1):
         
     | 
| 446 | 
         
             
                        input_file = gr.File(
         
     | 
| 447 | 
         
             
                            label="Upload Mass Spectrometry File (.mgf, .mzml, .mzxml)",
         
     | 
| 448 | 
         
            -
                            file_types=[".mgf", ".mzml", ".mzxml"]
         
     | 
| 449 | 
         
             
                        )
         
     | 
| 450 | 
         
             
                        decoding_method = gr.Radio(
         
     | 
| 451 | 
         
            -
                            [ 
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 452 | 
         
             
                            label="Decoding Method",
         
     | 
| 453 | 
         
            -
                            value="Greedy Search (Fast, resonably accurate)" 
     | 
| 454 | 
         
             
                        )
         
     | 
| 455 | 
         
             
                        submit_btn = gr.Button("Predict Sequences", variant="primary")
         
     | 
| 456 | 
         
             
                    with gr.Column(scale=2):
         
     | 
| 457 | 
         
            -
                        output_df = gr.DataFrame( 
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 458 | 
         
             
                        output_file = gr.File(label="Download Full Results (CSV)")
         
     | 
| 459 | 
         | 
| 460 | 
         
             
                submit_btn.click(
         
     | 
| 461 | 
         
             
                    predict_peptides,
         
     | 
| 462 | 
         
             
                    inputs=[input_file, decoding_method],
         
     | 
| 463 | 
         
            -
                    outputs=[output_df, output_file]
         
     | 
| 464 | 
         
             
                )
         
     | 
| 465 | 
         | 
| 466 | 
         
             
                gr.Examples(
         
     | 
| 467 | 
         
            -
             
     | 
| 468 | 
         
            -
             
     | 
| 469 | 
         
            -
             
     | 
| 470 | 
         
            -
             
     | 
| 471 | 
         
            -
             
     | 
| 472 | 
         
            -
             
     | 
| 473 | 
         
            -
             
     | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 474 | 
         
             
                )
         
     | 
| 475 | 
         | 
| 476 | 
         
             
                gr.Markdown(
         
     | 
| 477 | 
         
            -
             
     | 
| 478 | 
         
             
                     **Notes:**
         
     | 
| 479 | 
         
            -
                     *   Predictions are based on the [InstaNovo](https://github.com/instadeepai/InstaNovo) model  
     | 
| 480 | 
         
             
                     *   Knapsack Beam Search uses pre-calculated mass constraints and yields better results but takes longer.
         
     | 
| 481 | 
         
             
                     *   `delta_mass_ppm` shows the lowest absolute precursor mass error (in ppm) across potential isotopes (0-1 neutron).
         
     | 
| 482 | 
         
             
                     *   Ensure your input file format is correctly specified. Large files may take time to process.
         
     | 
| 
         @@ -487,10 +542,32 @@ with gr.Blocks(css=css, theme=gr.themes.Default(primary_hue="blue", secondary_hu 
     | 
|
| 487 | 
         
             
                with gr.Accordion("Application Logs", open=True):
         
     | 
| 488 | 
         
             
                    log_display = Log(log_file, dark=True, height=300)
         
     | 
| 489 | 
         | 
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 
         | 
|
| 490 | 
         
             
            # --- Launch the App ---
         
     | 
| 491 | 
         
             
            if __name__ == "__main__":
         
     | 
| 492 | 
         
             
                # Set share=True for temporary public link if running locally
         
     | 
| 493 | 
         
             
                # Set server_name="0.0.0.0" to allow access from network if needed
         
     | 
| 494 | 
         
             
                # demo.launch(server_name="0.0.0.0", server_port=7860)
         
     | 
| 495 | 
         
             
                # For Hugging Face Spaces, just demo.launch() is usually sufficient
         
     | 
| 496 | 
         
            -
                demo.launch(share=True) 
     | 
| 
         | 
|
| 28 | 
         
             
                raise ImportError(f"Failed to import InstaNovo components: {e}")
         
     | 
| 29 | 
         | 
| 30 | 
         
             
            # --- Configuration ---
         
     | 
| 31 | 
         
            +
            MODEL_ID = "instanovo-v1.1.0"  # Use the desired pretrained model ID
         
     | 
| 32 | 
         
             
            KNAPSACK_DIR = Path("./knapsack_cache")
         
     | 
| 33 | 
         
            +
            DEFAULT_CONFIG_PATH = Path(
         
     | 
| 34 | 
         
            +
                "./configs/inference/default.yaml"
         
     | 
| 35 | 
         
            +
            )  # Assuming instanovo installs configs locally relative to execution
         
     | 
| 36 | 
         | 
| 37 | 
         
             
            # Determine device
         
     | 
| 38 | 
         
             
            DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
         
     | 
| 39 | 
         
            +
            FP16 = DEVICE == "cuda"  # Enable FP16 only on CUDA
         
     | 
| 40 | 
         | 
| 41 | 
         
             
            # --- Global Variables (Load Model and Knapsack Once) ---
         
     | 
| 42 | 
         
             
            MODEL: InstaNovo | None = None
         
     | 
| 
         | 
|
| 80 | 
         | 
| 81 | 
         
             
                # --- Knapsack Handling ---
         
     | 
| 82 | 
         
             
                knapsack_exists = (
         
     | 
| 83 | 
         
            +
                    (KNAPSACK_DIR / "parameters.pkl").exists()
         
     | 
| 84 | 
         
            +
                    and (KNAPSACK_DIR / "masses.npy").exists()
         
     | 
| 85 | 
         
            +
                    and (KNAPSACK_DIR / "chart.npy").exists()
         
     | 
| 86 | 
         
             
                )
         
     | 
| 87 | 
         | 
| 88 | 
         
             
                if knapsack_exists:
         
     | 
| 
         | 
|
| 98 | 
         
             
                if not knapsack_exists:
         
     | 
| 99 | 
         
             
                    logger.info("Knapsack not found or failed to load. Generating knapsack...")
         
     | 
| 100 | 
         
             
                    if RESIDUE_SET is None:
         
     | 
| 101 | 
         
            +
                        raise gr.Error(
         
     | 
| 102 | 
         
            +
                            "Cannot generate knapsack because ResidueSet failed to load."
         
     | 
| 103 | 
         
            +
                        )
         
     | 
| 104 | 
         
             
                    try:
         
     | 
| 105 | 
         
             
                        # Prepare residue masses for knapsack generation (handle negative/zero masses)
         
     | 
| 106 | 
         
             
                        residue_masses_knapsack = dict(RESIDUE_SET.residue_masses.copy())
         
     | 
| 107 | 
         
            +
                        negative_residues = [
         
     | 
| 108 | 
         
            +
                            k for k, v in residue_masses_knapsack.items() if v <= 0
         
     | 
| 109 | 
         
            +
                        ]
         
     | 
| 110 | 
         
             
                        if negative_residues:
         
     | 
| 111 | 
         
             
                            logger.info(f"Warning: Non-positive masses found in residues: {negative_residues}. "
         
     | 
| 112 | 
         
             
                                  "Excluding from knapsack generation.")
         
     | 
| 
         | 
|
| 114 | 
         
             
                                del residue_masses_knapsack[res]
         
     | 
| 115 | 
         
             
                        # Remove special tokens explicitly if they somehow got mass
         
     | 
| 116 | 
         
             
                        for special_token in RESIDUE_SET.special_tokens:
         
     | 
| 117 | 
         
            +
                            if special_token in residue_masses_knapsack:
         
     | 
| 118 | 
         
            +
                                del residue_masses_knapsack[special_token]
         
     | 
| 119 | 
         | 
| 120 | 
         
             
                        # Ensure residue indices used match those without special/negative masses
         
     | 
| 121 | 
         
             
                        valid_residue_indices = {
         
     | 
| 122 | 
         
            +
                            res: idx
         
     | 
| 123 | 
         
            +
                            for res, idx in RESIDUE_SET.residue_to_index.items()
         
     | 
| 124 | 
         
             
                            if res in residue_masses_knapsack
         
     | 
| 125 | 
         
             
                        }
         
     | 
| 126 | 
         | 
| 
         | 
|
| 127 | 
         
             
                        KNAPSACK = Knapsack.construct_knapsack(
         
     | 
| 128 | 
         
             
                            residue_masses=residue_masses_knapsack,
         
     | 
| 129 | 
         
            +
                            residue_indices=valid_residue_indices,  # Use only valid indices
         
     | 
| 130 | 
         
             
                            max_mass=MAX_MASS,
         
     | 
| 131 | 
         
             
                            mass_scale=MASS_SCALE,
         
     | 
| 132 | 
         
             
                        )
         
     | 
| 
         | 
|
| 141 | 
         
             
            # Load the model and knapsack when the script starts
         
     | 
| 142 | 
         
             
            load_model_and_knapsack()
         
     | 
| 143 | 
         | 
| 144 | 
         
            +
             
     | 
| 145 | 
         
             
            def create_inference_config(
         
     | 
| 146 | 
         
             
                input_path: str,
         
     | 
| 147 | 
         
             
                output_path: str,
         
     | 
| 
         | 
|
| 150 | 
         
             
                """Creates the OmegaConf DictConfig needed for prediction."""
         
     | 
| 151 | 
         
             
                # Load default config if available, otherwise create from scratch
         
     | 
| 152 | 
         
             
                if DEFAULT_CONFIG_PATH.exists():
         
     | 
| 153 | 
         
            +
                    base_cfg = OmegaConf.load(DEFAULT_CONFIG_PATH)
         
     | 
| 154 | 
         
             
                else:
         
     | 
| 155 | 
         
             
                     logger.info(f"Warning: Default config not found at {DEFAULT_CONFIG_PATH}. Using minimal config.")
         
     | 
| 156 | 
         
             
                     # Create a minimal config if default is missing
         
     | 
| 
         | 
|
| 213 | 
         
             
                    cfg_overrides["use_knapsack"] = False
         
     | 
| 214 | 
         
             
                elif "Knapsack" in decoding_method:
         
     | 
| 215 | 
         
             
                    if KNAPSACK is None:
         
     | 
| 216 | 
         
            +
                        raise gr.Error(
         
     | 
| 217 | 
         
            +
                            "Knapsack is not available. Cannot use Knapsack Beam Search."
         
     | 
| 218 | 
         
            +
                        )
         
     | 
| 219 | 
         
             
                    cfg_overrides["num_beams"] = 5
         
     | 
| 220 | 
         
             
                    cfg_overrides["use_knapsack"] = True
         
     | 
| 221 | 
         
             
                    cfg_overrides["knapsack_path"] = str(KNAPSACK_DIR)
         
     | 
| 
         | 
|
| 232 | 
         
             
                Main function to load data, run prediction, and return results.
         
     | 
| 233 | 
         
             
                """
         
     | 
| 234 | 
         
             
                if MODEL is None or RESIDUE_SET is None or MODEL_CONFIG is None:
         
     | 
| 235 | 
         
            +
                    load_model_and_knapsack()  # Attempt to reload if None (e.g., after space restart)
         
     | 
| 236 | 
         
            +
                    if MODEL is None:
         
     | 
| 237 | 
         
            +
                        raise gr.Error("InstaNovo model is not loaded. Cannot perform prediction.")
         
     | 
| 238 | 
         | 
| 239 | 
         
             
                if input_file is None:
         
     | 
| 240 | 
         
             
                    raise gr.Error("Please upload a mass spectrometry file.")
         
     | 
| 
         | 
|
| 257 | 
         
             
                    try:
         
     | 
| 258 | 
         
             
                        sdf = SpectrumDataFrame.load(
         
     | 
| 259 | 
         
             
                            config.data_path,
         
     | 
| 260 | 
         
            +
                            lazy=False,  # Load eagerly for Gradio simplicity
         
     | 
| 261 | 
         
            +
                            is_annotated=False,  # De novo mode
         
     | 
| 262 | 
         
             
                            column_mapping=config.get("column_map", None),
         
     | 
| 263 | 
         
             
                            shuffle=False,
         
     | 
| 264 | 
         
            +
                            verbose=True,  # Print loading logs
         
     | 
| 265 | 
         
             
                        )
         
     | 
| 266 | 
         
             
                        # Apply charge filter like in CLI
         
     | 
| 267 | 
         
             
                        original_size = len(sdf)
         
     | 
| 268 | 
         
             
                        max_charge = config.get("max_charge", 10)
         
     | 
| 269 | 
         
             
                        sdf.filter_rows(
         
     | 
| 270 | 
         
            +
                            lambda row: (row["precursor_charge"] <= max_charge)
         
     | 
| 271 | 
         
            +
                            and (row["precursor_charge"] > 0)
         
     | 
| 272 | 
         
             
                        )
         
     | 
| 273 | 
         
             
                        if len(sdf) < original_size:
         
     | 
| 274 | 
         
             
                            logger.info(f"Warning: Filtered {original_size - len(sdf)} spectra with charge > {max_charge} or <= 0.")
         
     | 
| 
         | 
|
| 285 | 
         
             
                        sdf,
         
     | 
| 286 | 
         
             
                        RESIDUE_SET,
         
     | 
| 287 | 
         
             
                        MODEL_CONFIG.get("n_peaks", 200),
         
     | 
| 288 | 
         
            +
                        return_str=True,  # Needed for greedy/beam search targets later (though not used here)
         
     | 
| 289 | 
         
             
                        annotated=False,
         
     | 
| 290 | 
         
            +
                        pad_spectrum_max_length=config.get("compile_model", False)
         
     | 
| 291 | 
         
            +
                        or config.get("use_flash_attention", False),
         
     | 
| 292 | 
         
             
                        bin_spectra=config.get("conv_peak_encoder", False),
         
     | 
| 293 | 
         
             
                    )
         
     | 
| 294 | 
         
             
                    dl = DataLoader(
         
     | 
| 295 | 
         
             
                        ds,
         
     | 
| 296 | 
         
             
                        batch_size=config.batch_size,
         
     | 
| 297 | 
         
            +
                        num_workers=0,  # Required by SpectrumDataFrame
         
     | 
| 298 | 
         
            +
                        shuffle=False,  # Required by SpectrumDataFrame
         
     | 
| 299 | 
         
             
                        collate_fn=collate_batch,
         
     | 
| 300 | 
         
             
                    )
         
     | 
| 301 | 
         | 
| 
         | 
|
| 304 | 
         
             
                    decoder: Decoder
         
     | 
| 305 | 
         
             
                    if config.use_knapsack:
         
     | 
| 306 | 
         
             
                        if KNAPSACK is None:
         
     | 
| 307 | 
         
            +
                            # This check should ideally be earlier, but double-check
         
     | 
| 308 | 
         
            +
                            raise gr.Error(
         
     | 
| 309 | 
         
            +
                                "Knapsack is required for Knapsack Beam Search but is not available."
         
     | 
| 310 | 
         
            +
                            )
         
     | 
| 311 | 
         
             
                        # KnapsackBeamSearchDecoder doesn't directly load from path in this version?
         
     | 
| 312 | 
         
             
                        # We load Knapsack globally, so just pass it.
         
     | 
| 313 | 
         
             
                        # If it needed path: decoder = KnapsackBeamSearchDecoder.from_file(model=MODEL, path=config.knapsack_path)
         
     | 
| 
         | 
|
| 329 | 
         
             
                    # 5. Run Prediction Loop (Adapted from instanovo/transformer/predict.py)
         
     | 
| 330 | 
         
             
                    logger.info("Starting prediction...")
         
     | 
| 331 | 
         
             
                    start_time = time.time()
         
     | 
| 332 | 
         
            +
                    results_list: list[
         
     | 
| 333 | 
         
            +
                        ScoredSequence | list
         
     | 
| 334 | 
         
            +
                    ] = []  # Store ScoredSequence or empty list
         
     | 
| 335 | 
         | 
| 336 | 
         
             
                    for i, batch in enumerate(dl):
         
     | 
| 337 | 
         
            +
                        spectra, precursors, spectra_mask, _, _ = (
         
     | 
| 338 | 
         
            +
                            batch  # Ignore peptides/masks for de novo
         
     | 
| 339 | 
         
            +
                        )
         
     | 
| 340 | 
         
             
                        spectra = spectra.to(DEVICE)
         
     | 
| 341 | 
         
             
                        precursors = precursors.to(DEVICE)
         
     | 
| 342 | 
         
             
                        spectra_mask = spectra_mask.to(DEVICE)
         
     | 
| 343 | 
         | 
| 344 | 
         
            +
                        with (
         
     | 
| 345 | 
         
            +
                            torch.no_grad(),
         
     | 
| 346 | 
         
            +
                            torch.amp.autocast(DEVICE, dtype=torch.float16, enabled=FP16),
         
     | 
| 347 | 
         
            +
                        ):
         
     | 
| 348 | 
         
             
                            # Beam search decoder might return list[list[ScoredSequence]] if return_beam=True
         
     | 
| 349 | 
         
             
                            # Greedy decoder returns list[ScoredSequence]
         
     | 
| 350 | 
         
             
                            # KnapsackBeamSearchDecoder returns list[ScoredSequence] or list[list[ScoredSequence]]
         
     | 
| 
         | 
|
| 354 | 
         
             
                                beam_size=config.num_beams,
         
     | 
| 355 | 
         
             
                                max_length=config.max_length,
         
     | 
| 356 | 
         
             
                                # Knapsack/Beam Search specific params if needed
         
     | 
| 357 | 
         
            +
                                mass_tolerance=config.get("filter_precursor_ppm", 20)
         
     | 
| 358 | 
         
            +
                                * 1e-6,  # Convert ppm to relative
         
     | 
| 359 | 
         
            +
                                max_isotope=config.isotope_error_range[1]
         
     | 
| 360 | 
         
            +
                                if config.isotope_error_range
         
     | 
| 361 | 
         
            +
                                else 1,
         
     | 
| 362 | 
         
            +
                                return_beam=False,  # Only get the top prediction for simplicity
         
     | 
| 363 | 
         
             
                            )
         
     | 
| 364 | 
         
             
                        results_list.extend(batch_predictions) # Should be list[ScoredSequence] or list[list]
         
     | 
| 365 | 
         
             
                        logger.info(f"Processed batch {i+1}/{len(dl)}")
         
     | 
| 
         | 
|
| 372 | 
         
             
                    output_data = []
         
     | 
| 373 | 
         
             
                    # Use sdf index columns + prediction results
         
     | 
| 374 | 
         
             
                    index_cols = [col for col in config.index_columns if col in sdf.df.columns]
         
     | 
| 375 | 
         
            +
                    base_df_pd = sdf.df.select(index_cols).to_pandas()  # Get base info
         
     | 
| 376 | 
         | 
| 377 | 
         
             
                    metrics_calc = Metrics(RESIDUE_SET, config.isotope_error_range)
         
     | 
| 378 | 
         | 
| 379 | 
         
             
                    for i, res in enumerate(results_list):
         
     | 
| 380 | 
         
            +
                        row_data = base_df_pd.iloc[i].to_dict()  # Get corresponding input data
         
     | 
| 381 | 
         
             
                        if isinstance(res, ScoredSequence) and res.sequence:
         
     | 
| 382 | 
         
             
                            sequence_str = "".join(res.sequence)
         
     | 
| 383 | 
         
             
                            row_data["prediction"] = sequence_str
         
     | 
| 384 | 
         
             
                            row_data["log_probability"] = f"{res.sequence_log_probability:.4f}"
         
     | 
| 385 | 
         
             
                            # Use metrics to calculate delta mass ppm for the top prediction
         
     | 
| 386 | 
         
             
                            try:
         
     | 
| 387 | 
         
            +
                                _, delta_mass_list = metrics_calc.matches_precursor(
         
     | 
| 388 | 
         
            +
                                    res.sequence,
         
     | 
| 389 | 
         
            +
                                    row_data["precursor_mz"],
         
     | 
| 390 | 
         
            +
                                    row_data["precursor_charge"],
         
     | 
| 391 | 
         
            +
                                )
         
     | 
| 392 | 
         
            +
                                # Find the smallest absolute ppm error across isotopes
         
     | 
| 393 | 
         
            +
                                min_abs_ppm = (
         
     | 
| 394 | 
         
            +
                                    min(abs(p) for p in delta_mass_list)
         
     | 
| 395 | 
         
            +
                                    if delta_mass_list
         
     | 
| 396 | 
         
            +
                                    else float("nan")
         
     | 
| 397 | 
         
            +
                                )
         
     | 
| 398 | 
         
            +
                                row_data["delta_mass_ppm"] = f"{min_abs_ppm:.2f}"
         
     | 
| 399 | 
         
             
                            except Exception as e:
         
     | 
| 400 | 
         
             
                                 logger.info(f"Warning: Could not calculate delta mass for prediction {i}: {e}")
         
     | 
| 401 | 
         
             
                                 row_data["delta_mass_ppm"] = "N/A"
         
     | 
| 
         | 
|
| 409 | 
         
             
                    output_df = pl.DataFrame(output_data)
         
     | 
| 410 | 
         | 
| 411 | 
         
             
                    # Ensure specific columns are present and ordered
         
     | 
| 412 | 
         
            +
                    display_cols = [
         
     | 
| 413 | 
         
            +
                        "scan_number",
         
     | 
| 414 | 
         
            +
                        "precursor_mz",
         
     | 
| 415 | 
         
            +
                        "precursor_charge",
         
     | 
| 416 | 
         
            +
                        "prediction",
         
     | 
| 417 | 
         
            +
                        "log_probability",
         
     | 
| 418 | 
         
            +
                        "delta_mass_ppm",
         
     | 
| 419 | 
         
            +
                    ]
         
     | 
| 420 | 
         
             
                    final_display_cols = []
         
     | 
| 421 | 
         
             
                    for col in display_cols:
         
     | 
| 422 | 
         
             
                        if col in output_df.columns:
         
     | 
| 
         | 
|
| 431 | 
         | 
| 432 | 
         
             
                    output_df_display = output_df.select(final_display_cols)
         
     | 
| 433 | 
         | 
| 
         | 
|
| 434 | 
         
             
                    # 7. Save full results to CSV
         
     | 
| 435 | 
         
             
                    logger.info(f"Saving results to {output_csv_path}...")
         
     | 
| 436 | 
         
             
                    output_df.write_csv(output_csv_path)
         
     | 
| 
         | 
|
| 446 | 
         
             
                    # Re-raise as Gradio error
         
     | 
| 447 | 
         
             
                    raise gr.Error(f"Prediction failed: {e}")
         
     | 
| 448 | 
         | 
| 449 | 
         
            +
             
     | 
| 450 | 
         
             
            # --- Gradio Interface ---
         
     | 
| 451 | 
         
             
            css = """
         
     | 
| 452 | 
         
             
            .gradio-container { font-family: sans-serif; }
         
     | 
| 
         | 
|
| 456 | 
         
             
            .logo-container img { margin-bottom: 1rem; }
         
     | 
| 457 | 
         
             
            """
         
     | 
| 458 | 
         | 
| 459 | 
         
            +
            with gr.Blocks(
         
     | 
| 460 | 
         
            +
                css=css, theme=gr.themes.Default(primary_hue="blue", secondary_hue="blue")
         
     | 
| 461 | 
         
            +
            ) as demo:
         
     | 
| 462 | 
         
             
                # --- Logo Display ---
         
     | 
| 463 | 
         
             
                gr.Markdown(
         
     | 
| 464 | 
         
             
                    """
         
     | 
| 
         | 
|
| 466 | 
         
             
                      <img src='/gradio_api/file=assets/instanovo.svg' alt="InstaNovo Logo" width="300" style="display: block; margin: 0 auto;">
         
     | 
| 467 | 
         
             
                    </div>
         
     | 
| 468 | 
         
             
                    """,
         
     | 
| 469 | 
         
            +
                    elem_classes="logo-container",  # Optional class for CSS targeting
         
     | 
| 470 | 
         
             
                )
         
     | 
| 471 | 
         | 
| 472 | 
         
             
                # --- App Content ---
         
     | 
| 
         | 
|
| 481 | 
         
             
                    with gr.Column(scale=1):
         
     | 
| 482 | 
         
             
                        input_file = gr.File(
         
     | 
| 483 | 
         
             
                            label="Upload Mass Spectrometry File (.mgf, .mzml, .mzxml)",
         
     | 
| 484 | 
         
            +
                            file_types=[".mgf", ".mzml", ".mzxml"],
         
     | 
| 485 | 
         
             
                        )
         
     | 
| 486 | 
         
             
                        decoding_method = gr.Radio(
         
     | 
| 487 | 
         
            +
                            [
         
     | 
| 488 | 
         
            +
                                "Greedy Search (Fast, resonably accurate)",
         
     | 
| 489 | 
         
            +
                                "Knapsack Beam Search (More accurate, but slower)",
         
     | 
| 490 | 
         
            +
                            ],
         
     | 
| 491 | 
         
             
                            label="Decoding Method",
         
     | 
| 492 | 
         
            +
                            value="Greedy Search (Fast, resonably accurate)",  # Default to fast method
         
     | 
| 493 | 
         
             
                        )
         
     | 
| 494 | 
         
             
                        submit_btn = gr.Button("Predict Sequences", variant="primary")
         
     | 
| 495 | 
         
             
                    with gr.Column(scale=2):
         
     | 
| 496 | 
         
            +
                        output_df = gr.DataFrame(
         
     | 
| 497 | 
         
            +
                            label="Prediction Results",
         
     | 
| 498 | 
         
            +
                            headers=[
         
     | 
| 499 | 
         
            +
                                "scan_number",
         
     | 
| 500 | 
         
            +
                                "precursor_mz",
         
     | 
| 501 | 
         
            +
                                "precursor_charge",
         
     | 
| 502 | 
         
            +
                                "prediction",
         
     | 
| 503 | 
         
            +
                                "log_probability",
         
     | 
| 504 | 
         
            +
                                "delta_mass_ppm",
         
     | 
| 505 | 
         
            +
                            ],
         
     | 
| 506 | 
         
            +
                            wrap=True,
         
     | 
| 507 | 
         
            +
                        )
         
     | 
| 508 | 
         
             
                        output_file = gr.File(label="Download Full Results (CSV)")
         
     | 
| 509 | 
         | 
| 510 | 
         
             
                submit_btn.click(
         
     | 
| 511 | 
         
             
                    predict_peptides,
         
     | 
| 512 | 
         
             
                    inputs=[input_file, decoding_method],
         
     | 
| 513 | 
         
            +
                    outputs=[output_df, output_file],
         
     | 
| 514 | 
         
             
                )
         
     | 
| 515 | 
         | 
| 516 | 
         
             
                gr.Examples(
         
     | 
| 517 | 
         
            +
                    [
         
     | 
| 518 | 
         
            +
                        ["assets/sample_spectra.mgf", "Greedy Search (Fast, resonably accurate)"],
         
     | 
| 519 | 
         
            +
                        [
         
     | 
| 520 | 
         
            +
                            "assets/sample_spectra.mgf",
         
     | 
| 521 | 
         
            +
                            "Knapsack Beam Search (More accurate, but slower)",
         
     | 
| 522 | 
         
            +
                        ],
         
     | 
| 523 | 
         
            +
                    ],
         
     | 
| 524 | 
         
            +
                    inputs=[input_file, decoding_method],
         
     | 
| 525 | 
         
            +
                    outputs=[output_df, output_file],
         
     | 
| 526 | 
         
            +
                    fn=predict_peptides,
         
     | 
| 527 | 
         
            +
                    cache_examples=False,  # Re-run examples if needed
         
     | 
| 528 | 
         
            +
                    label="Example Usage",
         
     | 
| 529 | 
         
             
                )
         
     | 
| 530 | 
         | 
| 531 | 
         
             
                gr.Markdown(
         
     | 
| 532 | 
         
            +
                    """
         
     | 
| 533 | 
         
             
                     **Notes:**
         
     | 
| 534 | 
         
            +
                     *   Predictions are based on the [InstaNovo](https://github.com/instadeepai/InstaNovo) model `{MODEL_ID}`.
         
     | 
| 535 | 
         
             
                     *   Knapsack Beam Search uses pre-calculated mass constraints and yields better results but takes longer.
         
     | 
| 536 | 
         
             
                     *   `delta_mass_ppm` shows the lowest absolute precursor mass error (in ppm) across potential isotopes (0-1 neutron).
         
     | 
| 537 | 
         
             
                     *   Ensure your input file format is correctly specified. Large files may take time to process.
         
     | 
| 
         | 
|
| 542 | 
         
             
                with gr.Accordion("Application Logs", open=True):
         
     | 
| 543 | 
         
             
                    log_display = Log(log_file, dark=True, height=300)
         
     | 
| 544 | 
         | 
| 545 | 
         
            +
                gr.Textbox(
         
     | 
| 546 | 
         
            +
                    value="""
         
     | 
| 547 | 
         
            +
            @article{eloff_kalogeropoulos_2025_instanovo,
         
     | 
| 548 | 
         
            +
            	title        = {InstaNovo enables diffusion-powered de novo peptide sequencing in large-scale proteomics experiments},
         
     | 
| 549 | 
         
            +
            	author       = {Kevin Eloff and Konstantinos Kalogeropoulos and Amandla Mabona and Oliver Morell and Rachel Catzel and 
         
     | 
| 550 | 
         
            +
                                Esperanza Rivera-de-Torre and Jakob Berg Jespersen and Wesley Williams and Sam P. B. van Beljouw and 
         
     | 
| 551 | 
         
            +
                                Marcin J. Skwark and Andreas Hougaard Laustsen and Stan J. J. Brouns and Anne Ljungars and Erwin M. 
         
     | 
| 552 | 
         
            +
                                Schoof and Jeroen Van Goey and Ulrich auf dem Keller and Karim Beguir and Nicolas Lopez Carranza and 
         
     | 
| 553 | 
         
            +
                                Timothy P. Jenkins},
         
     | 
| 554 | 
         
            +
            	year         = 2025,
         
     | 
| 555 | 
         
            +
            	month        = {Mar},
         
     | 
| 556 | 
         
            +
            	day          = 31,
         
     | 
| 557 | 
         
            +
            	journal      = {Nature Machine Intelligence},
         
     | 
| 558 | 
         
            +
            	doi          = {10.1038/s42256-025-01019-5},
         
     | 
| 559 | 
         
            +
            	url          = {https://www.nature.com/articles/s42256-025-01019-5}
         
     | 
| 560 | 
         
            +
            }
         
     | 
| 561 | 
         
            +
            """,
         
     | 
| 562 | 
         
            +
                    show_copy_button=True,
         
     | 
| 563 | 
         
            +
                    label="If you use InstaNovo in your research, please cite:",
         
     | 
| 564 | 
         
            +
                    interactive=False,
         
     | 
| 565 | 
         
            +
                )
         
     | 
| 566 | 
         
            +
             
     | 
| 567 | 
         
             
            # --- Launch the App ---
         
     | 
| 568 | 
         
             
            if __name__ == "__main__":
         
     | 
| 569 | 
         
             
                # Set share=True for temporary public link if running locally
         
     | 
| 570 | 
         
             
                # Set server_name="0.0.0.0" to allow access from network if needed
         
     | 
| 571 | 
         
             
                # demo.launch(server_name="0.0.0.0", server_port=7860)
         
     | 
| 572 | 
         
             
                # For Hugging Face Spaces, just demo.launch() is usually sufficient
         
     | 
| 573 | 
         
            +
                demo.launch(share=True)  # For local testing with public URL
         
     |