SERENE

This model is a fine-tuned version of facebook/wav2vec2-base on the mali111222333/SERENE_DATA dataset. It achieves the following results on the evaluation set:

• Loss: 0.5125
• Accuracy: 0.8262

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 0
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• 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
2.001	1.0	103	2.0125	0.3141
1.4204	2.0	206	1.5098	0.5219
1.1139	3.0	309	1.1529	0.6792
0.9586	4.0	412	0.9849	0.7157
0.9062	5.0	515	0.9828	0.6956
0.8295	6.0	618	0.8643	0.7236
0.7713	7.0	721	0.7047	0.7722
0.7446	8.0	824	0.6894	0.7861
0.7021	9.0	927	0.6706	0.7837
0.7081	10.0	1030	0.6263	0.8026
0.646	11.0	1133	0.5994	0.7977
0.6489	12.0	1236	0.5866	0.8026
0.6152	13.0	1339	0.5655	0.8074
0.6346	14.0	1442	0.5525	0.8202
0.6076	15.0	1545	0.5299	0.8250
0.5924	16.0	1648	0.5210	0.8238
0.6078	17.0	1751	0.5254	0.8232
0.5703	18.0	1854	0.5128	0.8232
0.6421	19.0	1957	0.5197	0.8220
0.5085	20.0	2060	0.5125	0.8262

Framework versions

• Transformers 4.57.0.dev0
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1