TrimLesson3

This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.0961
• Accuracy: 0.7003
• F1-score: 0.6957
• Recall-score: 0.7003
• Precision-score: 0.7014

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: 20
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1-score	Recall-score	Precision-score
4.0066	1.0	2068	3.5593	0.1583	0.0819	0.1583	0.0794
2.5652	2.0	4136	2.2228	0.4437	0.3810	0.4437	0.4189
1.9375	3.0	6204	1.5382	0.5597	0.5157	0.5597	0.5544
2.1447	4.0	8272	1.3384	0.6030	0.5647	0.6030	0.5980
2.1308	5.0	10340	1.2420	0.6216	0.5906	0.6216	0.6206
1.7815	6.0	12408	1.1685	0.6384	0.6109	0.6384	0.6326
1.1674	7.0	14476	1.1605	0.6431	0.6197	0.6431	0.6433
1.5469	8.0	16544	1.1038	0.6674	0.6420	0.6674	0.6617
0.6686	9.0	18612	1.0640	0.6708	0.6494	0.6708	0.6588
1.2668	10.0	20680	1.1181	0.6669	0.6457	0.6669	0.6564
0.5084	11.0	22748	1.0662	0.6773	0.6597	0.6773	0.6770
1.7345	12.0	24816	1.0945	0.6783	0.6641	0.6783	0.6821
0.7144	13.0	26884	1.0492	0.6857	0.6715	0.6857	0.6903
0.712	14.0	28952	1.0526	0.6900	0.6791	0.6900	0.6940
2.2976	15.0	31020	1.0654	0.6960	0.6847	0.6960	0.7023
0.6391	16.0	33088	1.0770	0.6912	0.6817	0.6912	0.6929
0.9704	17.0	35156	1.0885	0.6949	0.6895	0.6949	0.7022
0.9055	18.0	37224	1.0743	0.6965	0.6916	0.6965	0.6987
2.0981	19.0	39292	1.0877	0.7025	0.6977	0.7025	0.7051
0.3026	20.0	41360	1.0961	0.7003	0.6957	0.7003	0.7014

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu118
• Datasets 2.20.0
• Tokenizers 0.20.0