Update gui.py

gui.py

@@ -11,6 +11,28 @@ import librosa
 import soundfile as sf
 from ensemble import ensemble_files
 import shutil
+import gradio_client.utils as client_utils
+
+# Patch gradio_client.utils.get_type to handle boolean schemas
+def patched_get_type(schema):
+    if isinstance(schema, bool):
+        return "boolean"
+    if "const" in schema:
+        return repr(schema["const"])
+    if "enum" in schema:
+        return f"Union[{', '.join(repr(e) for e in schema['enum'])}]"
+    if "type" not in schema:
+        return "Any"
+    type_ = schema["type"]
+    if isinstance(type_, list):
+        return f"Union[{', '.join(t for t in type_ if t != 'null')}]"
+    if type_ == "array":
+        return f"List[{client_utils._json_schema_to_python_type(schema.get('items', {}), schema.get('$defs', {}))}]"
+    if type_ == "object":
+        return "Dict[str, Any]"
+    return type_
+
+client_utils.get_type = patched_get_type
 
 # Device and autocast setup
 device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -132,8 +154,7 @@ CSS = """
     overflow: hidden;
 }
 body {
-    background: url('/content/logo.jpg') no-repeat center center fixed;
-    background-size: cover;
+    background: none;
     margin: 0;
     padding: 0;
     font-family: 'Roboto', sans-serif;
@@ -335,7 +356,6 @@ def download_audio(url, out_dir="ytdl"):
     if not url:
         raise ValueError("No URL provided.")
     
-    # Clear ytdl directory
     if os.path.exists(out_dir):
         shutil.rmtree(out_dir)
     os.makedirs(out_dir, exist_ok=True)
@@ -358,10 +378,8 @@ def roformer_separator(audio, model_key, seg_size, override_seg_size, overlap, p
     if not audio:
         raise ValueError("No audio file provided.")
     
-    # Convert override_seg_size to boolean
     override_seg_size = override_seg_size == "True"
     
-    # Clear output directory
     if os.path.exists(output_dir):
         shutil.rmtree(output_dir)
     os.makedirs(output_dir, exist_ok=True)
@@ -392,7 +410,6 @@ def roformer_separator(audio, model_key, seg_size, override_seg_size, overlap, p
         separation = separator.separate(audio)
         stems = [os.path.join(output_dir, file_name) for file_name in separation]
         
-        # Filter excluded stems
         if exclude_stems.strip():
             excluded = [s.strip().lower() for s in exclude_stems.split(',')]
             filtered_stems = [stem for stem in stems if not any(ex in os.path.basename(stem).lower() for ex in excluded)]
@@ -402,15 +419,13 @@ def roformer_separator(audio, model_key, seg_size, override_seg_size, overlap, p
         logger.error(f"Separation failed: {e}")
         raise RuntimeError(f"Separation failed: {e}")
 
-def auto_ensemble_process(audio, model_keys, seg_size, overlap, out_format, use_tta, model_dir, output_dir, norm_thresh, amp_thresh, batch_size, ensemble_method, exclude_stems="", weights=None, progress=gr.Progress()):
+def auto_ensemble_process(audio, model_keys, seg_size, overlap, out_format, use_tta, model_dir, output_dir, norm_thresh, amp_thresh, batch_size, ensemble_method, exclude_stems="", weights_str="", progress=gr.Progress()):
     """Perform ensemble processing on audio using multiple Roformer models."""
     if not audio or not model_keys:
         raise ValueError("Audio or models missing.")
     
-    # Convert use_tta to boolean
     use_tta = use_tta == "True"
     
-    # Clear output directory
     if os.path.exists(output_dir):
         shutil.rmtree(output_dir)
     os.makedirs(output_dir, exist_ok=True)
@@ -421,7 +436,6 @@ def auto_ensemble_process(audio, model_keys, seg_size, overlap, out_format, use_
     all_stems = []
     total_models = len(model_keys)
     
-    # Separate audio with each model
     for i, model_key in enumerate(model_keys):
         for category, models in ROFORMER_MODELS.items():
             if model_key in models:
@@ -446,7 +460,6 @@ def auto_ensemble_process(audio, model_keys, seg_size, overlap, out_format, use_
         separation = separator.separate(audio)
         stems = [os.path.join(output_dir, file_name) for file_name in separation]
         
-        # Filter excluded stems
         if exclude_stems.strip():
             excluded = [s.strip().lower() for s in exclude_stems.split(',')]
             filtered_stems = [stem for stem in stems if not any(ex in os.path.basename(stem).lower() for ex in excluded)]
@@ -457,11 +470,10 @@ def auto_ensemble_process(audio, model_keys, seg_size, overlap, out_format, use_
     if not all_stems:
         raise ValueError("No valid stems for ensemble after exclusion.")
     
-    # Default weights if none provided
-    if weights is None or len(weights) != len(all_stems):
+    weights = [float(w.strip()) for w in weights_str.split(',')] if weights_str.strip() else [1.0] * len(all_stems)
+    if len(weights) != len(all_stems):
         weights = [1.0] * len(all_stems)
     
-    # Perform ensemble
     output_file = os.path.join(output_dir, f"{base_name}_ensemble_{ensemble_method}.{out_format}")
     ensemble_args = [
         "--files", *all_stems,
@@ -477,11 +489,11 @@ def auto_ensemble_process(audio, model_keys, seg_size, overlap, out_format, use_
 
 def update_roformer_models(category):
     """Update Roformer model dropdown based on selected category."""
-    return gr.update(choices=list(ROFORMER_MODELS.get(category, {}).keys()))
+    return gr.update(choices=list(ROFORMER_MODELS.get(category, {}).keys()) or [])
 
 def update_ensemble_models(category):
     """Update ensemble model dropdown based on selected category."""
-    return gr.update(choices=list(ROFORMER_MODELS.get(category, {}).keys()))
+    return gr.update(choices=list(ROFORMER_MODELS.get(category, {}).keys()) or [])
 
 # Interface creation
 def create_interface():
@@ -497,8 +509,8 @@ def create_interface():
                     model_file_dir = gr.Textbox(value="/tmp/audio-separator-models/", label="📂 Model Cache", placeholder="Path to model directory", interactive=True)
                     output_dir = gr.Textbox(value="output", label="📤 Output Directory", placeholder="Where to save results", interactive=True)
                     output_format = gr.Dropdown(value="wav", choices=OUTPUT_FORMATS, label="🎶 Output Format", interactive=True)
-                    norm_threshold = gr.Slider(0.1, 1, value=0.9, step=0.1, label="🔊 Normalization Threshold", interactive=True)
-                    amp_threshold = gr.Slider(0.1, 1, value=0.3, step=0.1, label="📈 Amplification Threshold", interactive=True)
+                    norm_threshold = gr.Slider(0.1, 1.0, value=0.9, step=0.1, label="🔊 Normalization Threshold", interactive=True)
+                    amp_threshold = gr.Slider(0.1, 1.0, value=0.3, step=0.1, label="📈 Amplification Threshold", interactive=True)
                     batch_size = gr.Slider(1, 16, value=4, step=1, label="⚡ Batch Size", interactive=True)
 
             # Roformer Tab
@@ -563,13 +575,7 @@ def create_interface():
         ensemble_category.change(update_ensemble_models, inputs=[ensemble_category], outputs=[ensemble_models])
         download_ensemble.click(fn=download_audio, inputs=[url_ensemble], outputs=[ensemble_audio])
         ensemble_button.click(
-            fn=lambda audio, models, seg_size, overlap, out_format, use_tta, model_dir, output_dir,
-                    norm_thresh, amp_thresh, batch_size, method, exclude_stems, weights_str:
-                auto_ensemble_process(
-                    audio, models, seg_size, overlap, out_format, use_tta, model_dir, output_dir,
-                    norm_thresh, amp_thresh, batch_size, method, exclude_stems,
-                    [float(w.strip()) for w in weights_str.split(',')] if weights_str.strip() else None
-                ),
+            fn=auto_ensemble_process,
             inputs=[
                 ensemble_audio, ensemble_models, ensemble_seg_size, ensemble_overlap,
                 output_format, ensemble_use_tta, model_file_dir, output_dir,
@@ -588,8 +594,8 @@ if __name__ == "__main__":
 
     app = create_interface()
     try:
-        # Use share=True for remote access or server_name="127.0.0.1" for local testing
-        app.launch(server_name="0.0.0.0", server_port=args.port, share=True)
+        # For Hugging Face Spaces or local testing
+        app.launch(server_name="0.0.0.0", server_port=args.port, share=False)
     except Exception as e:
         logger.error(f"Failed to launch app: {e}")
         raise