Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Upload folder using huggingface_hub

Reduce audio row height (compact-audio CSS); sync dataset_loader

Results: add Results (CLAP) and Results (Dasheng) views; dataset_loader supports dasheng results

Speed up: cache repo file list; Dasheng dropdown exclude fold*

Dasheng view: dropdown only shows batch_outputs_dasheng sample IDs (no fold*)

View options: Nearest Neighbor (CLAP) and Nearest Neighbor (Dasheng); dataset_loader supports batch_outputs_dasheng

NN view: NN1-NN10 from baseline (prompt order), no separate baseline block

Add view toggle: Nearest Neighbor (baseline + 10 NN) vs Results (3 baselines + Ours, top 3 prompts)

Restructure: per-prompt blocks with Baseline then Ours (1 image + 3 audios each)

Show retrieved prompt for each NN1/NN2/NN3

Redesign UI: single noise selector, then Baseline/NN1/NN2/NN3 each with 1 image + 3 audios (BG/FG/Mix); add get_noise_demo_paths, no on-the-fly spectrogram

Add Gradio app: NN demo + comparison, dataset_loader, audio_utils, requirements