jbilcke-hf's picture
jbilcke-hf HF Staff
fixed ui persistence issues
ab45a2c
"""
Hugging Face Hub tab for Video Model Studio UI.
Handles browsing, searching, and importing datasets from the Hugging Face Hub.
"""
import gradio as gr
import logging
import asyncio
import threading
from pathlib import Path
from typing import Dict, Any, List, Optional, Tuple
from vms.utils import BaseTab
logger = logging.getLogger(__name__)
class HubTab(BaseTab):
"""Hub tab for importing datasets from Hugging Face Hub"""
def __init__(self, app_state):
super().__init__(app_state)
self.id = "hub_tab"
self.title = "Import from Hugging Face"
self.is_downloading = False
def create(self, parent=None) -> gr.Tab:
"""Create the Hub tab UI components"""
with gr.Tab(self.title, id=self.id) as tab:
with gr.Column():
with gr.Row():
gr.Markdown("## Import a dataset from Hugging Face")
with gr.Row():
with gr.Column():
with gr.Row():
gr.Markdown("You can use any dataset containing video files (.mp4) with optional captions (same names but in .txt format)")
with gr.Row():
gr.Markdown("You can also use a dataset containing WebDataset shards (.tar files).")
with gr.Column():
self.components["dataset_search"] = gr.Textbox(
label="Search Hugging Face Datasets (MP4, WebDataset)",
placeholder="video datasets eg. cakeify, disney, rickroll.."
)
with gr.Row():
self.components["dataset_search_btn"] = gr.Button(
"Search Datasets",
variant="primary",
#size="md"
)
# Dataset browser results section
with gr.Row(visible=False) as dataset_results_row:
self.components["dataset_results_row"] = dataset_results_row
with gr.Column(scale=3):
self.components["dataset_results"] = gr.Dataframe(
headers=["Dataset ID"], # Simplified to show only dataset ID
interactive=False,
wrap=True,
row_count=10,
label="Dataset Results"
)
with gr.Column(scale=3):
# Dataset info and state
self.components["dataset_info"] = gr.Markdown("Select a dataset to see details")
self.components["dataset_id"] = gr.State(value=None)
self.components["file_type"] = gr.State(value=None)
self.components["download_in_progress"] = gr.State(value=False)
# Files section that appears when a dataset is selected
with gr.Column(visible=False) as files_section:
self.components["files_section"] = files_section
# Video files row (appears if videos are present)
with gr.Row() as video_files_row:
self.components["video_files_row"] = video_files_row
self.components["video_count_text"] = gr.Markdown("Contains 0 video files")
self.components["download_videos_btn"] = gr.Button("Download", variant="primary")
# WebDataset files row (appears if tar files are present)
with gr.Row() as webdataset_files_row:
self.components["webdataset_files_row"] = webdataset_files_row
self.components["webdataset_count_text"] = gr.Markdown("Contains 0 WebDataset (.tar) files")
self.components["download_webdataset_btn"] = gr.Button("Download", variant="primary")
# Status indicator
self.components["status_output"] = gr.Markdown("")
return tab
def connect_events(self) -> None:
"""Connect event handlers to UI components"""
# Dataset search event
self.components["dataset_search_btn"].click(
fn=self.search_datasets,
inputs=[self.components["dataset_search"]],
outputs=[
self.components["dataset_results"],
self.components["dataset_results_row"]
]
)
# Dataset selection event
self.components["dataset_results"].select(
fn=self.display_dataset_info,
outputs=[
self.components["dataset_info"],
self.components["dataset_id"],
self.components["files_section"],
self.components["video_files_row"],
self.components["video_count_text"],
self.components["webdataset_files_row"],
self.components["webdataset_count_text"],
self.components["status_output"] # Reset status output
]
)
# Check if we have access to project_tabs_component
if hasattr(self.app, "project_tabs_component"):
tabs_component = self.app.project_tabs_component
else:
# Fallback to prevent errors
logger.warning("project_tabs_component not found in app, using None for tab switching")
tabs_component = None
# Download videos button
self.components["download_videos_btn"].click(
fn=self.set_file_type_and_return,
outputs=[self.components["file_type"]]
).then(
fn=self.download_file_group,
inputs=[
self.components["dataset_id"],
self.components["enable_automatic_video_split"],
self.components["file_type"]
],
outputs=[
self.components["status_output"],
self.components["import_status"],
self.components["download_videos_btn"],
self.components["download_webdataset_btn"],
self.components["download_in_progress"]
]
).success(
fn=self.app.tabs["import_tab"].on_import_success,
inputs=[
self.components["enable_automatic_video_split"],
self.components["enable_automatic_content_captioning"],
self.app.tabs["caption_tab"].components["custom_prompt_prefix"]
],
outputs=[
tabs_component,
self.components["status_output"]
]
)
# Download WebDataset button
self.components["download_webdataset_btn"].click(
fn=self.set_file_type_and_return_webdataset,
outputs=[self.components["file_type"]]
).then(
fn=self.download_file_group,
inputs=[
self.components["dataset_id"],
self.components["enable_automatic_video_split"],
self.components["file_type"]
],
outputs=[
self.components["status_output"],
self.components["import_status"],
self.components["download_videos_btn"],
self.components["download_webdataset_btn"],
self.components["download_in_progress"]
]
).success(
fn=self.app.tabs["import_tab"].on_import_success,
inputs=[
self.components["enable_automatic_video_split"],
self.components["enable_automatic_content_captioning"],
self.app.tabs["caption_tab"].components["custom_prompt_prefix"]
],
outputs=[
tabs_component,
self.components["status_output"]
]
)
def set_file_type_and_return(self):
"""Set file type to video and return it"""
return "video"
def set_file_type_and_return_webdataset(self):
"""Set file type to webdataset and return it"""
return "webdataset"
def search_datasets(self, query: str):
"""Search datasets on the Hub matching the query"""
try:
logger.info(f"Searching for datasets with query: '{query}'")
results_full = self.app.importing.search_datasets(query)
# Extract just the first column (dataset IDs) for display
results = [[row[0]] for row in results_full]
return results, gr.update(visible=True)
except Exception as e:
logger.error(f"Error searching datasets: {str(e)}", exc_info=True)
return [[f"Error: {str(e)}"]], gr.update(visible=True)
def display_dataset_info(self, evt: gr.SelectData):
"""Display detailed information about the selected dataset"""
try:
if not evt or not evt.value:
logger.warning("No dataset selected in display_dataset_info")
return (
"No dataset selected", # dataset_info
None, # dataset_id
gr.update(visible=False), # files_section
gr.update(visible=False), # video_files_row
"", # video_count_text
gr.update(visible=False), # webdataset_files_row
"", # webdataset_count_text
"" # status_output
)
# Extract dataset_id from the simplified format
dataset_id = evt.value[0] if isinstance(evt.value, list) else evt.value
logger.info(f"Getting dataset info for: {dataset_id}")
# Use the importer service to get dataset info
info_text, file_counts, _ = self.app.importing.get_dataset_info(dataset_id)
# Get counts of each file type
video_count = file_counts.get("video", 0)
webdataset_count = file_counts.get("webdataset", 0)
# Return all the required outputs individually
return (
info_text, # dataset_info
dataset_id, # dataset_id
gr.update(visible=True), # files_section
gr.update(visible=video_count > 0), # video_files_row
f"Contains {video_count} video file{'s' if video_count != 1 else ''}", # video_count_text
gr.update(visible=webdataset_count > 0), # webdataset_files_row
f"Contains {webdataset_count} WebDataset (.tar) file{'s' if webdataset_count != 1 else ''}", # webdataset_count_text
"" # status_output
)
except Exception as e:
logger.error(f"Error displaying dataset info: {str(e)}", exc_info=True)
return (
f"Error loading dataset information: {str(e)}", # dataset_info
None, # dataset_id
gr.update(visible=False), # files_section
gr.update(visible=False), # video_files_row
"", # video_count_text
gr.update(visible=False), # webdataset_files_row
"", # webdataset_count_text
"" # status_output
)
async def _download_with_progress(self, dataset_id, file_type, enable_splitting, progress_callback):
"""Wrapper for download_file_group that integrates with progress tracking"""
try:
# Set up the progress callback adapter
def progress_adapter(progress_value, desc=None, total=None):
# For a progress bar, we need to convert the values to a 0-1 range
if isinstance(progress_value, (int, float)):
if total is not None and total > 0:
# If we have a total, calculate the fraction
fraction = min(1.0, progress_value / total)
else:
# Otherwise, just use the value directly (assumed to be 0-1)
fraction = min(1.0, progress_value)
# Update the progress with the calculated fraction
progress_callback(fraction, desc=desc)
# Call the actual download function with our adapter
result = await self.app.importing.download_file_group(
dataset_id,
file_type,
enable_splitting,
progress_callback=progress_adapter
)
return result
except Exception as e:
logger.error(f"Error in download with progress: {str(e)}", exc_info=True)
return f"Error: {str(e)}"
def download_file_group(self, dataset_id: str, enable_splitting: bool, file_type: str, progress=gr.Progress()) -> Tuple:
"""Handle download of a group of files (videos or WebDatasets) with progress tracking"""
try:
if not dataset_id:
return ("No dataset selected",
"No dataset selected",
gr.update(),
gr.update(),
False)
logger.info(f"Starting download of {file_type} files from dataset: {dataset_id}")
# Initialize progress tracking
progress(0, desc=f"Starting download of {file_type} files from {dataset_id}")
# Disable download buttons during the process
videos_btn_update = gr.update(interactive=False)
webdataset_btn_update = gr.update(interactive=False)
# Run the download function with progress tracking
# We need to use asyncio.run to run the coroutine in a synchronous context
result = asyncio.run(self._download_with_progress(
dataset_id,
file_type,
enable_splitting,
progress
))
# When download is complete, update the UI
progress(1.0, desc="Download complete!")
# Create a success message
success_msg = f"✅ Download complete! {result}"
# Update the UI components
return (
success_msg, # status_output - shows the successful result
result, # import_status
gr.update(interactive=True), # download_videos_btn
gr.update(interactive=True), # download_webdataset_btn
False # download_in_progress
)
except Exception as e:
error_msg = f"Error downloading {file_type} files: {str(e)}"
logger.error(error_msg, exc_info=True)
return (
f"❌ Error: {error_msg}", # status_output
error_msg, # import_status
gr.update(interactive=True), # download_videos_btn
gr.update(interactive=True), # download_webdataset_btn
False # download_in_progress
)