Update app.py

app.py

@@ -3,17 +3,13 @@ import torchaudio
 from sgmse.model import ScoreModel
 import gradio as gr
 from sgmse.util.other import pad_spec
-
 # Load the pre-trained model
-model = SGMSENoiseReducer.from_pretrained("sp-uhh/speech-enhancement-sgmse")
-
+model = ScoreModel.load_from_checkpoint("https://huggingface.co/sp-uhh/speech-enhancement-sgmse/resolve/main/train_vb_29nqe0uh_epoch%3D115.ckpt")
 def enhance_speech(audio_file):
     # Load and process the audio file
     y, sr = torchaudio.load(audio_file)
-
-    T_orig = y.size(1)   
-
-    # Normalize
+T_orig = y.size(1)   
+# Normalize
     norm_factor = y.abs().max()
     y = y / norm_factor
     
@@ -29,8 +25,7 @@ def enhance_speech(audio_file):
     
     # Backward transform in time domain
     x_hat = model.to_audio(sample.squeeze(), T_orig)
-
-    # Renormalize
+# Renormalize
     x_hat = x_hat * norm_factor
     
     # Save the enhanced audio
@@ -38,13 +33,12 @@ def enhance_speech(audio_file):
     torchaudio.save(output_file, x_hat.cpu().numpy(), sr)
     
     return output_file
-
 # Gradio interface setup
 inputs = gr.Audio(label="Input Audio", type="filepath")
 outputs = gr.Audio(label="Output Audio", type="filepath")
 title = "Speech Enhancement using SGMSE"
 description = "This Gradio demo uses the SGMSE model for speech enhancement. Upload your audio file to enhance it."
 article = "<p style='text-align: center'><a href='https://huggingface.co/SP-UHH/speech-enhancement-sgmse' target='_blank'>Model Card</a></p>"
-
 # Launch without share=True (as it's not supported on Hugging Face Spaces)
-gr.Interface(fn=enhance_speech, inputs=inputs, outputs=outputs, title=title, description=description, article=article).launch()
+gr.Interface(fn=enhance_speech, inputs=inputs, outputs=outputs, title=title, description=description, article=article).launch(
+