debug the configuration error

app.py

@@ -1,44 +1,37 @@
-import torch
-import torchaudio
 import gradio as gr
 import spaces  # Enables ZeroGPU on Hugging Face
-from demucs import pretrained
-from demucs.apply import apply_model
+from transformers import AutoModelForCausalLM
+from anticipation.sample import generate
+from anticipation.convert import events_to_midi, midi_to_events
+from anticipation import ops
+from anticipation.tokenize import extract_instruments
+import torch
 from pyharp import *
-from audiotools import AudioSignal
+from safetensors.torch import load_file
+import os
 
 
-# Available Demucs models
 DEMUX_MODELS = ["mdx_extra_q", "mdx_extra", "htdemucs", "mdx_q"]
 
 STEM_CHOICES = {
-    "Vocals": 3,   
-    "Drums": 0,    
-    "Bass": 1,     
-    "Other": 2,    
+    "Vocals": 3,
+    "Drums": 0,
+    "Bass": 1,
+    "Other": 2,
     "Instrumental (No Vocals)": "instrumental"
 }
 
-@spaces.GPU
-def separate_stem(audio_file_path: str, model_name: str, stem_choice: str):
-    """
-    Separates an audio file into the chosen stem using a Demucs model.
-    Ensures correct stem ordering and supports mono input.
-    """
-    # Load Demucs model
+@space.GPU
+def separate_stem(audio_file_path: str, model_name: str, stem_choice: str) -> AudioSignal:
     model = pretrained.get_model(model_name)
     model.to('cuda' if torch.cuda.is_available() else 'cpu')
     model.eval()
 
-    # Load the audio file
     waveform, sr = torchaudio.load(audio_file_path)
-
-    # Check if input is mono
     is_mono = waveform.shape[0] == 1
     if is_mono:
-        waveform = waveform.repeat(2, 1)  # Convert mono to stereo for Demucs
+        waveform = waveform.repeat(2, 1)
 
-    # Apply Demucs model
     with torch.no_grad():
         stems_batch = apply_model(
             model,
@@ -48,24 +41,20 @@ def separate_stem(audio_file_path: str, model_name: str, stem_choice: str):
             split=True
         )
 
-    # stems shape: (batch, stems, channels, samples)
-    stems = stems_batch[0]  
-
-    print(f"Model '{model_name}' extracted stems shape: {stems.shape}")
+    stems = stems_batch[0]
 
     if stem_choice == "Instrumental (No Vocals)":
-        stem = stems[0] + stems[1] + stems[2]  # Drums + Bass + Other 
+        stem = stems[0] + stems[1] + stems[2]
     else:
         stem_index = STEM_CHOICES[stem_choice]
         stem = stems[stem_index]
 
-    # Convert back to mono if the input was originally mono
     if is_mono:
-        stem = stem.mean(dim=0, keepdim=True)  # Stereo → Mono
+        stem = stem.mean(dim=0, keepdim=True)
 
-    # Convert to AudioSignal with float32 dtype
-    stem_signal = AudioSignal(stem.cpu().numpy().astype('float32'), sample_rate=sr)
-    return stem_signal
+    return AudioSignal(stem.cpu().numpy().astype('float32'), sample_rate=sr)
+
+# Gradio Callback Function
 
 def process_fn_stem(audio_file_path: str, demucs_model: str, stem_choice: str):
     """
@@ -78,7 +67,7 @@ def process_fn_stem(audio_file_path: str, demucs_model: str, stem_choice: str):
     return stem_path, LabelList(labels=[])
 
 
-# Define the model card
+# Model Card 
 model_card = ModelCard(
     name="Demucs Stem Separator",
     description="Uses Demucs to separate a music track into a selected stem.",
@@ -86,9 +75,9 @@ model_card = ModelCard(
     tags=["demucs", "source-separation", "pyharp", "stems"]
 )
 
-# Build Gradio interface with dropdowns for model and stem selection
+# Gradio UI
 with gr.Blocks() as demo:
-    gr.LoginButton()
+
     dropdown_model = gr.Dropdown(
         label="Select Demucs Model",
         choices=DEMUX_MODELS,

requirements.txt

@@ -1,6 +1,5 @@
-
 -e git+https://github.com/TEAMuP-dev/pyharp.git#egg=pyharp
-demucs==4.0.0
+demucs
 dora-search
 einops
 julius>=0.2.3
@@ -8,11 +7,13 @@ lameenc>=1.2
 openunmix
 pyyaml
 tqdm
-torch==2.0.1
+torch>=1.8.1, <2.1
 torchaudio>=0.8, <2.1
 diffq>=0.2.1  
 ffmpeg  
 numpy<2
 scipy  
 soundfile  
-hydra-core>=1.1 
+hydra-core>=1.1  
+typing
+pydantic==2.10.6