wav2vec2-gtzan

Fine-tuned facebook/wav2vec2-base for GTZAN music-genre classification (10 classes). Trained with 15s-window augmentation on sanchit-gandhi/gtzan; song-level accuracy 0.92 (window 0.8894). Use with AutoModelForAudioClassification.

Metrics

accuracy: 0.92

Usage

# audio-classification

Downloads last month: -

Safetensors

Model size

94.6M params

Tensor type

F32