hubert-base-ls960-speech-commands

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

• Loss: 1.0829
• Accuracy: 0.8058

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: 48
• eval_batch_size: 48
• seed: 42
• 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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.8285	1.0	824	1.9509	0.7167
0.5292	2.0	1648	1.3813	0.7909
0.3554	3.0	2472	1.1773	0.7941
0.2873	4.0	3296	1.2437	0.7981
0.2525	5.0	4120	1.2514	0.8004
0.2941	6.0	4944	1.2243	0.7995
0.1809	7.0	5768	1.1965	0.8008
0.2313	8.0	6592	1.0694	0.8022
0.1917	9.0	7416	1.0618	0.7995
0.1212	10.0	8240	1.0972	0.8026
0.185	11.0	9064	1.0868	0.8017
0.143	12.0	9888	1.1558	0.8031
0.2227	13.0	10712	1.0550	0.8040
0.1884	14.0	11536	1.0384	0.8022
0.1183	15.0	12360	1.0169	0.8035
0.1849	16.0	13184	1.0061	0.8035
0.141	17.0	14008	1.0337	0.8053
0.1328	18.0	14832	1.0829	0.8058
0.1238	19.0	15656	1.0576	0.8053
0.0932	20.0	16480	1.0641	0.8053

Framework versions

• Transformers 4.43.3
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.19.1