wav2vec2transformerEMR3

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: 0.6589
• Accuracy: 0.7916
• Precision: 0.7918
• Recall: 0.7916
• F1: 0.7896

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
1.4706	1.6420	500	1.4207	0.5409	0.5559	0.5409	0.5180
0.953	3.2841	1000	0.9292	0.7275	0.7438	0.7275	0.7233
0.7575	4.9261	1500	0.7618	0.7616	0.7686	0.7616	0.7610
0.6084	6.5681	2000	0.7485	0.7559	0.7658	0.7559	0.7524
0.5221	8.2102	2500	0.6990	0.7711	0.7767	0.7711	0.7691
0.431	9.8522	3000	0.6967	0.7752	0.7796	0.7752	0.7719
0.3814	11.4943	3500	0.6523	0.7867	0.7875	0.7867	0.7856
0.3461	13.1363	4000	0.6589	0.7916	0.7918	0.7916	0.7896
0.3405	14.7783	4500	0.6703	0.7867	0.7878	0.7867	0.7847

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Tokenizers 0.20.3