Upload folder using huggingface_hub

library.csv

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2e32e0f5bdb86dd24f0f7531469b106315fe68914fc48b427f0d859e733c2493
-size 88173
+oid sha256:ed730a77eb0d3d1ffd3c94288bb944014280dbdd587d9ae3f4331f0a908544a2
+size 346166

sync_library_and_hf.py

@@ -2,13 +2,7 @@
 # sync_library_and_hf.py
 ''''
 RUN BELOW FOR NEW HTML FILES TO UPDATE OLD ONES ON DFATASET REPO
-python sync_library_and_hf.py  
---db-path library.csv  
- --repo-id akazemian/audio-html 
- --model-name wavcoch_audio-preds-sr=16000   
- --index-filename index.csv   
- --wipe-remote   
- --wipe-local
+python sync_library_and_hf.py --db-path library.csv  --repo-id akazemian/audio-html --model-name wavcoch_audio-preds-sr=16000   --index-filename index.csv   --wipe-remote   --wipe-local
 '''
 import argparse, datetime, uuid, posixpath, sys, traceback, os, hashlib
 from pathlib import Path
@@ -16,6 +10,7 @@ from typing import List, Tuple, Set
 from urllib.parse import unquote
 import os
 import pandas as pd
+import numpy as np
 from huggingface_hub import (
     HfApi,
     hf_hub_download,
@@ -57,6 +52,12 @@ def create_file_specific_manifest(csv_path: Path) -> pd.DataFrame:
     audio_dir = "/data/atlask/BAU-Quant/val"
     manifest = pd.read_csv(csv_path)
 
+    mask = manifest['dataset'].eq('TUT_urban_acoustic_scenes')
+    manifest['audio_category'] = np.where(mask, manifest['dataset'], manifest['audio_category'])
+    manifest = manifest.assign(
+        audio_category = manifest['audio_category'].where(~mask, manifest['dataset'])
+    )
+
     # 1) Build a files dataframe
     files = pd.DataFrame({"file_name": os.listdir(audio_dir)})
     # keep only audio files if needed