Update app.py

app.py

@@ -1,77 +1,110 @@
+
 import gradio as gr
+import os
+from datasets import load_dataset, Dataset
 import pandas as pd
-from datasets import load_dataset, Dataset, Audio
 from huggingface_hub import login
 
-# Global variables to store dataset and token
-editable_df = pd.DataFrame()
-dataset_name = ""
-hub_token = ""
-
-
-def load_hf_dataset(dataset_url, token):
-    global editable_df, dataset_name, hub_token
-    # Extract dataset name and token
-    dataset_name = dataset_url.split("/")[-1]
-    hub_token = token
+def load_huggingface_dataset(dataset_link, token):
+    """
+    Load a Hugging Face dataset using the provided link and token.
+    """
+    # Extract dataset name and config if applicable
+    parts = dataset_link.split("/")
+    dataset_name = parts[-1]
+    if len(parts) > 2:
+        owner = parts[-2]
+        dataset_name = f"{owner}/{dataset_name}"
+    
+    # Load the dataset
+    dataset = load_dataset(dataset_link, split="train")
+    
+    # Return the dataset as a DataFrame with index and transcription columns
+    df = dataset.to_pandas().reset_index()
+    return df[["index", "transcription"]], dataset
 
-    # Authenticate and load dataset
+def update_transcriptions(df, dataset, token,dataset_link):
+    """
+    Update the transcriptions in the dataset and push it back to the Hugging Face Hub.
+    """
+    # Convert DataFrame back to Dataset
+    updated_dataset = Dataset.from_pandas(df)
+    
+    # print(updated_dataset)
+    # print(dataset)
+    
+    # Replace the original transcription column in the dataset
+    dataset = dataset.map(
+        lambda examples, idx: {"transcription": updated_dataset["transcription"][idx]},
+        with_indices=True 
+    )
+    print(dataset['transcription'][0])
+    
     login(token)
-    dataset = load_dataset(dataset_url)
-    editable_df = pd.DataFrame(dataset["train"])
-    return editable_df
-
-def update_row(row_index, column_name, new_value):
-    """Update a specific cell in the DataFrame."""
-    global editable_df
-    if row_index < len(editable_df) and column_name in editable_df.columns:
-        editable_df.at[row_index, column_name] = new_value.replace('"', '')
-        print(new_value.replace('"', ''))
-    return editable_df
-
-def save_and_upload():
-    """Save the updated DataFrame back to the Hugging Face Hub."""
-    global editable_df, dataset_name, hub_token
-    # Convert DataFrame to Dataset
-    updated_dataset = Dataset.from_pandas(editable_df)
-
-    updated_dataset = updated_dataset.cast_column("audio", Audio(sampling_rate=16000))
-
-    # Push updated dataset to Hugging Face
-    updated_dataset.push_to_hub(dataset_name, token=hub_token)
-    return f"Updated dataset successfully pushed to: {dataset_name}"
-
-def handle_row_selection(selected_row, evt: gr.SelectData):
-    # print(selected_row)
-    index = evt.index[0]
-    return index , selected_row.transcription[index]
-
-# Gradio interface
-with gr.Blocks() as app:
-    gr.Markdown("### Hugging Face Dataset Editor")
     
-    with gr.Row():
-        dataset_url_input = gr.Textbox(label="Dataset URL", placeholder="username/dataset_name")
-        token_input = gr.Textbox(label="Hub Token", placeholder="Enter your Hugging Face Hub token", type="password")
-        load_btn = gr.Button("Load Dataset")
-
-    data_table = gr.DataFrame(value=editable_df)
+    dataset.push_to_hub(dataset_link)
     
-    with gr.Row():
-        row_input = gr.Number(label="Row Index", value=0, precision=0, interactive=False)
-        col_input = gr.Text(label="Column Name", value="transcription", interactive=False)
-        new_value_input = gr.Text(label="New Value", value="new_value", interactive=True)
-        update_btn = gr.Button("Update Row")
+    return "Dataset updated and changes submitted to the Hugging Face Hub!"
 
-    # Register callback to handle row selection and update
-    data_table.select(handle_row_selection, data_table,[row_input,new_value_input])#
+# Gradio Interface
+def main():
+    
     
-    save_btn = gr.Button("Save and Upload")
-    status_output = gr.Textbox(label="Status", interactive=False)
+    dataset = None  # To store the loaded dataset object globally
+    
+    def load_dataset_and_show_table(dataset_link, token):
+        """
+        Load the dataset and return the DataFrame to display in Gradio.
+        """
+        nonlocal dataset
+        df, dataset = load_huggingface_dataset(dataset_link, token)
+        return df
+    
+    def submit_changes(df, token,dataset_link):
+        """
+        Submit updated changes to the Hugging Face Hub.
+        """
+        if dataset is None:
+            return "No dataset loaded to update."
+        
+        print(df)
+        print(token)
+        return update_transcriptions(df, dataset, token,dataset_link)
 
-    # Button actions
-    load_btn.click(load_hf_dataset, inputs=[dataset_url_input, token_input], outputs=data_table)
-    update_btn.click(update_row, inputs=[row_input, col_input, new_value_input], outputs=data_table)
-    save_btn.click(save_and_upload, outputs=status_output)
+    # Gradio Interface
+    with gr.Blocks(css=".dataframe-row { height: 200px; }") as interface:
+        gr.Markdown("## Hugging Face Audio Dataset Editor")
+        
+        # Input fields for dataset link and token
+        dataset_link = gr.Textbox(label="Hugging Face Dataset Link")
+        hf_token = gr.Textbox(label="Hugging Face Token", type="password")
+        
+        # Button to load dataset
+        load_button = gr.Button("Load Dataset")
+        
+        # Table to display and edit dataset
+        table = gr.Dataframe(
+            headers=["index", "transcription"],
+            datatype=["number", "str"],
+            interactive=True,
+            label="Edit Dataset (Transcriptions are RTL)",
+        )
+        
+        # Button to submit changes
+        submit_button = gr.Button("Submit Changes")
+        output_message = gr.Textbox(label="Message")
+        
+        # RTL styling for transcription column
+        table.style = {"transcription": {"direction": "rtl"}}
+        
+        # Button functionality
+        load_button.click(load_dataset_and_show_table, [dataset_link, hf_token], table)
+        submit_button.click(submit_changes, [table, hf_token,dataset_link], output_message)
+    
+    # Launch Gradio Interface
+    interface.launch(share=True)
+    
+    
 
-app.launch(share=True)
+if __name__ == "__main__":
+    main()