mitro99/distilhubert-finetuned-gtzan_batch4_grad16_cosinelr

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:

• Accuracy: 0.85
• Loss: 0.7531

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
2.2849	1.0	14	0.17	2.2588
2.1931	1.99	28	0.47	2.0874
1.9194	2.99	42	0.58	1.8044
1.6351	3.98	56	0.61	1.5806
1.4473	4.98	70	0.71	1.3886
1.3131	5.97	84	0.7	1.2738
1.2141	6.97	98	0.72	1.1616
1.0657	7.96	112	0.74	1.1272
0.96	8.96	126	0.75	1.0251
0.8387	9.96	140	0.8	0.9364
0.8653	10.95	154	0.79	0.8858
0.7653	11.95	168	0.8	0.8233
0.7329	12.94	182	0.83	0.7982
0.675	13.94	196	0.81	0.8189
0.6174	14.93	210	0.82	0.8236
0.5714	16.0	225	0.82	0.7755
0.598	17.0	239	0.81	0.7511
0.5794	17.99	253	0.84	0.7553
0.589	18.99	267	0.85	0.7533
0.5717	19.91	280	0.85	0.7531

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0