trillsson3-ft-keyword-spotting-11

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: 0.3166
• Accuracy: 0.9088

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: 32
• eval_batch_size: 32
• seed: 0
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• 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
2.9219	1.0	399	1.2023	0.6217
0.9604	2.0	798	0.5437	0.8117
0.7608	3.0	1197	0.4222	0.8888
0.7045	4.0	1596	0.3881	0.8932
0.659	5.0	1995	0.3706	0.8847
0.6541	6.0	2394	0.3553	0.8917
0.6448	7.0	2793	0.3482	0.8953
0.6288	8.0	3192	0.3409	0.8989
0.641	9.0	3591	0.3297	0.9051
0.6369	10.0	3990	0.3325	0.9042
0.6218	11.0	4389	0.3250	0.9064
0.6247	12.0	4788	0.3312	0.8959
0.6284	13.0	5187	0.3217	0.9069
0.6213	14.0	5586	0.3301	0.8978
0.6274	15.0	5985	0.3180	0.9081
0.627	16.0	6384	0.3257	0.9020
0.6227	17.0	6783	0.3193	0.9056
0.6192	18.0	7182	0.3199	0.9066
0.6075	19.0	7581	0.3183	0.9073
0.6196	20.0	7980	0.3166	0.9088

Framework versions

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