Update app.py

app.py

@@ -6,7 +6,6 @@ DATASET_NAME = "Cnam-LMSSC/vibravox-test"
 SUBSETS = ["speech_clean", "speech_noisy", "speechless_clean", "speechless_noisy"]
 SPLITS = ["train", "validation", "test"]
 TEXT_COLUMN = "raw_text"
-# Add new column names to the configuration
 PHONEMIZED_TEXT_COLUMN = "phonemized_text"
 GENDER_COLUMN = "gender"
 AUDIO_COLUMNS = [
@@ -28,23 +27,30 @@ def load_and_update_all(subset, split):
         dataset = load_dataset(DATASET_NAME, name=subset, split=split)
         has_text_fields = TEXT_COLUMN in dataset.features
 
-        # Get the first row to display immediately
         sample = dataset[0]
-        sentence = sample[TEXT_COLUMN] if has_text_fields else None
-        # Fetch the new fields
-        phonemized_text = sample[PHONEMIZED_TEXT_COLUMN] if has_text_fields else None
-        gender = sample[GENDER_COLUMN] if has_text_fields else None
+        sentence = sample.get(TEXT_COLUMN)
+        phonemized_text = sample.get(PHONEMIZED_TEXT_COLUMN)
+        gender = sample.get(GENDER_COLUMN)
         
         raw_audio_data = [
             (sample[col]['sampling_rate'], sample[col]['array']) for col in AUDIO_COLUMNS
         ]
         
-        # Return updates for all UI components, including the new ones
+        # --- THE FIX IS HERE ---
+        # We add a condition to handle datasets with only one row.
+        dataset_len = len(dataset)
+        if dataset_len <= 1:
+            # If there's only one item, hide the slider as it's not needed.
+            slider_update = gr.update(visible=False)
+        else:
+            # Otherwise, show and configure the slider as normal.
+            slider_update = gr.update(maximum=dataset_len - 1, value=0, visible=True, interactive=True)
+        # --------------------
+
         return (
             dataset,
-            gr.update(maximum=len(dataset) - 1, value=0, visible=True, interactive=True),
+            slider_update, # Use the new slider_update variable here
             gr.update(value=sentence, visible=has_text_fields),
-            # Add updates for the new text boxes
             gr.update(value=phonemized_text, visible=has_text_fields),
             gr.update(value=gender, visible=has_text_fields),
             *raw_audio_data,
@@ -53,7 +59,6 @@ def load_and_update_all(subset, split):
     except Exception as e:
         error_message = f"Failed to load {subset}/{split}. Error: {e}"
         empty_audio = (None, None)
-        # Return empty/hidden updates for all components on error
         return (
             None,
             gr.update(visible=False),
@@ -67,16 +72,15 @@ def get_audio_row(dataset, index):
     Fetches a new row from the currently loaded dataset when the slider moves.
     """
     if dataset is None:
-        return [None] * (3 + len(AUDIO_COLUMNS)) # 3 text fields now
+        return [None] * (3 + len(AUDIO_COLUMNS))
 
     index = int(index)
     sample = dataset[index]
 
     has_text_fields = TEXT_COLUMN in dataset.features
-    sentence = sample[TEXT_COLUMN] if has_text_fields else None
-    # Fetch the new fields for the selected row
-    phonemized_text = sample[PHONEMIZED_TEXT_COLUMN] if has_text_fields else None
-    gender = sample[GENDER_COLUMN] if has_text_fields else None
+    sentence = sample.get(TEXT_COLUMN)
+    phonemized_text = sample.get(PHONEMIZED_TEXT_COLUMN)
+    gender = sample.get(GENDER_COLUMN)
 
     raw_audio_data = [
         (sample[col]['sampling_rate'], sample[col]['array']) for col in AUDIO_COLUMNS
@@ -84,26 +88,19 @@ def get_audio_row(dataset, index):
 
     return [sentence, phonemized_text, gender] + raw_audio_data
 
+# --- Build the Gradio Interface (No changes needed here) ---
 with gr.Blocks(css="footer {display: none !important}") as demo:
-    # Change the app title
     gr.Markdown("# Vibravox Viewer")
-
     loaded_dataset_state = gr.State(None)
-
     with gr.Row():
         subset_dropdown = gr.Dropdown(SUBSETS, value="speech_clean", label="Select Subset")
         split_dropdown = gr.Dropdown(SPLITS, value="train", label="Select Split")
-    
     error_box = gr.Textbox(visible=False, interactive=False, container=False)
-    
-    # Group the text outputs together
     with gr.Row():
         sentence_output = gr.Textbox(label="Raw Text", interactive=False)
         phonemized_output = gr.Textbox(label="Phonemized Text", interactive=False)
         gender_output = gr.Textbox(label="Gender", interactive=False)
-
     slider = gr.Slider(label="Select Data Row")
-    
     with gr.Row():
         audio1 = gr.Audio(label="Headset Microphone")
         audio2 = gr.Audio(label="Laryngophone (Throat Mic)")
@@ -113,21 +110,12 @@ with gr.Blocks(css="footer {display: none !important}") as demo:
         audio5 = gr.Audio(label="Forehead Accelerometer")
         audio6 = gr.Audio(label="Temple Vibration Pickup")
 
-    # Update the component lists to include the new text boxes
     all_outputs = [loaded_dataset_state, slider, sentence_output, phonemized_output, gender_output, audio1, audio2, audio3, audio4, audio5, audio6, error_box]
     data_outputs = [sentence_output, phonemized_output, gender_output, audio1, audio2, audio3, audio4, audio5, audio6]
-
-    # --- WIRING THE EVENT HANDLERS ---
-    # The handlers themselves don't need to change, as we updated the functions and component lists
     
-    # 1. When the app first loads
     demo.load(fn=load_and_update_all, inputs=[subset_dropdown, split_dropdown], outputs=all_outputs)
-    
-    # 2. When a dropdown value changes
     subset_dropdown.change(fn=load_and_update_all, inputs=[subset_dropdown, split_dropdown], outputs=all_outputs)
     split_dropdown.change(fn=load_and_update_all, inputs=[subset_dropdown, split_dropdown], outputs=all_outputs)
-
-    # 3. When ONLY the slider changes
     slider.change(fn=get_audio_row, inputs=[loaded_dataset_state, slider], outputs=data_outputs)
 
 demo.launch()