Upload README.md with huggingface_hub

README.md

@@ -1,10 +1,4 @@
 ---
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: "data/train-*.parquet"
-  default: true
 language:
 - ha
 license: mit
@@ -58,20 +52,22 @@ Each example contains:
 
 ## Usage
 
-### Basic Loading
+### Basic Loading (Recommended)
+
+The dataset will automatically download audio files when you access them:
 
 ```python
 from datasets import load_dataset, Audio
 
-# Load dataset
+# Load dataset - this downloads the parquet file
 dataset = load_dataset("Aybee5/HausaTTSEmbed", split="train")
 
-# IMPORTANT: Cast audio column to Audio type to enable audio loading
+# Cast audio column - this tells HF to download audio files when accessed
 dataset = dataset.cast_column("audio", Audio(sampling_rate=22050))
 
 print(f"Loaded {len(dataset)} samples")
 
-# Access first sample
+# Access first sample - audio file is downloaded automatically
 sample = dataset[0]
 print(f"Text: {sample['text']}")
 print(f"Speaker: {sample['speaker_id']}")
@@ -79,6 +75,8 @@ print(f"Audio shape: {sample['audio']['array'].shape}")
 print(f"Sample rate: {sample['audio']['sampling_rate']}")
 ```
 
+**Note:** Audio files are downloaded on-demand when you access them. For batch processing, they'll be cached locally.
+
 ### For Unsloth TTS Training
 
 This dataset is optimized for Unsloth TTS training:
@@ -98,7 +96,7 @@ if "source" not in raw_ds.column_names and "speaker_id" not in raw_ds.column_nam
 elif "source" not in raw_ds.column_names and "speaker_id" in raw_ds.column_names:
     speaker_key = "speaker_id"
 
-# Resample to target sample rate
+# Resample to target sample rate - downloads audio files as needed
 target_sampling_rate = 24000
 raw_ds = raw_ds.cast_column("audio", Audio(sampling_rate=target_sampling_rate))
 
@@ -110,6 +108,8 @@ val_ds = split_ds['test']
 print(f"Train: {len(train_ds)}, Validation: {len(val_ds)}")
 ```
 
+**Important for Colab/Remote:** Audio files are automatically downloaded from HuggingFace when accessed. The first access will download and cache the files.
+
 ### With Transformers
 
 ```python