hubert-esc50-finetuned

This model is a fine-tuned version of facebook/hubert-base-ls960 on the ESC-50 dataset. It achieves the following results on the evaluation set:

• Loss: 1.0816
• Accuracy: 0.8325

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.5937	1.0	200	3.4961	0.1
3.1597	2.0	400	3.1798	0.1325
2.8922	3.0	600	2.8387	0.2025
2.6376	4.0	800	2.5594	0.285
2.1292	5.0	1000	2.3671	0.35
2.1607	6.0	1200	2.0533	0.4225
1.7886	7.0	1400	1.8790	0.42
1.626	8.0	1600	1.7147	0.52
1.5246	9.0	1800	1.6021	0.545
0.9318	10.0	2000	1.4441	0.5825
0.9384	11.0	2200	1.2180	0.67
0.9081	12.0	2400	1.1540	0.7075
0.803	13.0	2600	1.1317	0.72
0.4613	14.0	2800	1.0722	0.74
0.4389	15.0	3000	1.1055	0.73
0.4175	16.0	3200	1.0409	0.725
0.2977	17.0	3400	0.9540	0.78
0.3455	18.0	3600	0.9743	0.805
0.2237	19.0	3800	1.0938	0.7775
0.154	20.0	4000	1.0646	0.8
0.0966	21.0	4200	1.0621	0.7875
0.172	22.0	4400	1.1815	0.7725
0.055	23.0	4600	1.1436	0.79
0.1465	24.0	4800	1.1070	0.81
0.0458	25.0	5000	1.1053	0.82
0.0137	26.0	5200	1.0798	0.815
0.0449	27.0	5400	1.1108	0.8225
0.0231	28.0	5600	1.1113	0.83
0.0218	29.0	5800	1.0896	0.83
0.047	30.0	6000	1.0816	0.8325

Framework versions

• Transformers 4.37.2
• Pytorch 2.0.1+cu117
• Datasets 2.16.1
• Tokenizers 0.15.1