trillsson3-ft-keyword-spotting-12

This model is a fine-tuned version of vumichien/nonsemantic-speech-trillsson3 on the superb dataset. It achieves the following results on the evaluation set:

• Loss: 0.3015
• Accuracy: 0.9150

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 64
• seed: 0
• 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: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2824	1.0	1597	0.7818	0.6892
0.8003	2.0	3194	0.4443	0.8735
0.7232	3.0	4791	0.3728	0.8833
0.73	4.0	6388	0.3465	0.8973
0.7015	5.0	7985	0.3211	0.9109
0.6981	6.0	9582	0.3200	0.9081
0.6807	7.0	11179	0.3209	0.9059
0.6873	8.0	12776	0.3206	0.9022
0.6416	9.0	14373	0.3124	0.9057
0.6698	10.0	15970	0.3288	0.8950
0.716	11.0	17567	0.3147	0.8998
0.6514	12.0	19164	0.3034	0.9112
0.6513	13.0	20761	0.3091	0.9092
0.652	14.0	22358	0.3056	0.9100
0.7105	15.0	23955	0.3015	0.9150
0.6337	16.0	25552	0.3070	0.9091
0.63	17.0	27149	0.3018	0.9135
0.6672	18.0	28746	0.3084	0.9088
0.6479	19.0	30343	0.3060	0.9101
0.6658	20.0	31940	0.3072	0.9089

Framework versions

• Transformers 4.23.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.6.1
• Tokenizers 0.13.1