distilhubert-finetuned-gtzan

This model is a fine-tuned version of ntu-spml/distilhubert on the GTZAN dataset. It achieves the following results on the evaluation set:

• Loss: 0.6228
• Accuracy: 0.85

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 4e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.2841	0.98	28	2.2578	0.22
2.108	2.0	57	2.0031	0.55
1.7117	2.98	85	1.6220	0.65
1.4624	4.0	114	1.4061	0.7
1.2607	4.98	142	1.1969	0.69
1.1162	6.0	171	1.0955	0.75
1.0	6.98	199	0.9670	0.78
0.8864	8.0	228	0.9192	0.77
0.8583	8.98	256	0.8475	0.78
0.8147	10.0	285	0.8214	0.77
0.6572	10.98	313	0.7754	0.78
0.5958	12.0	342	0.7187	0.79
0.4196	12.98	370	0.6732	0.83
0.4515	14.0	399	0.7272	0.8
0.4256	14.98	427	0.6507	0.84
0.3734	16.0	456	0.6587	0.83
0.3541	16.98	484	0.6244	0.86
0.312	18.0	513	0.6363	0.84
0.3287	18.98	541	0.6226	0.86
0.313	19.65	560	0.6228	0.85

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3