update app for file scp setup

app.py

@@ -277,8 +277,8 @@ def evaluate_audio(gt_file, pred_file, metric_config, include_timestamps=False):
     if not versa_installed:
         return None, "VERSA is not installed. Evaluation cannot be performed."
     
-    if gt_file is None or pred_file is None:
-        return None, "Please upload both ground truth and prediction audio files."
+    if pred_file is None:
+        return None, "Please upload the audio file to be evaluated."
     
     # Determine the metric config path
     if metric_config.startswith("custom/"):
@@ -291,13 +291,26 @@ def evaluate_audio(gt_file, pred_file, metric_config, include_timestamps=False):
     # Create temp directory for results
     with tempfile.TemporaryDirectory() as temp_dir:
         output_file = os.path.join(temp_dir, "result.json")
-        
+
+        # Create SCP file
+        pred_scp_path = os.path.join(temp_dir, "pred.scp")
+        with open(pred_scp_path, "w") as pred_scp:
+            pred_scp.write("test {}\n".format(pred_file))
+       
+        # For case without reference audio
+        if gt_file is not None:
+            gt_scp_path = os.path.join(temp_dir, "gt.scp")
+            with open(gt_scp_path, "w") as gt_scp:
+                gt_scp.write("test {}\n".format(gt_file))
+        else:
+            gt_scp_path = "None"
+
         # Build command
         cmd = [
             sys.executable, VERSA_BIN,
             "--score_config", metric_config_path,
-            "--gt", gt_file,
-            "--pred", pred_file,
+            "--gt", gt_scp_path,
+            "--pred", pred_scp_path,
             "--output_file", output_file,
             "--use_gpu", "true",
         ]
@@ -378,8 +391,8 @@ def create_gradio_demo():
                 with gr.TabItem("Standard Evaluation"):
                     with gr.Row():
                         with gr.Column():
-                            gt_audio = gr.Audio(label="Ground Truth Audio", type="filepath", sources=["upload", "microphone"])
                             pred_audio = gr.Audio(label="Prediction Audio", type="filepath", sources=["upload", "microphone"])
+                            gt_audio = gr.Audio(label="Ground Truth Audio", type="filepath", sources=["upload", "microphone"])
                             
                             metric_dropdown = gr.Dropdown(
                                 choices=available_metrics,