trillsson3-ft-keyword-spotting-13

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: 0.3093
• Accuracy: 0.9153

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: 32
• seed: 0
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• 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
1.7756	1.0	798	0.9283	0.6396
0.8631	2.0	1596	0.4884	0.8573
0.7551	3.0	2394	0.3967	0.8832
0.6968	4.0	3192	0.3644	0.8989
0.67	5.0	3990	0.3428	0.9057
0.6854	6.0	4788	0.3408	0.9026
0.6701	7.0	5586	0.3359	0.9013
0.6734	8.0	6384	0.3285	0.9059
0.6581	9.0	7182	0.3199	0.9095
0.6557	10.0	7980	0.3301	0.8986
0.6768	11.0	8778	0.3174	0.9047
0.6459	12.0	9576	0.3192	0.9031
0.6607	13.0	10374	0.3173	0.9066
0.656	14.0	11172	0.3142	0.9094
0.6302	15.0	11970	0.3093	0.9153
0.636	16.0	12768	0.3184	0.9044
0.6327	17.0	13566	0.3104	0.9117
0.6428	18.0	14364	0.3158	0.9084
0.6515	19.0	15162	0.3129	0.9097
0.6441	20.0	15960	0.3115	0.9100

Framework versions

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