finetuned_wav2vec2.0-base-on-IEMOCAP_2

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: 1.1569
• Accuracy: 0.7390

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.1881	0.99	112	1.2005	0.4768
1.0121	2.0	225	1.0271	0.5619
0.8569	3.0	338	0.9382	0.6018
0.8679	4.0	451	0.8015	0.6947
0.5643	4.99	563	0.7752	0.7046
0.4579	6.0	676	0.7699	0.7400
0.3993	7.0	789	0.8323	0.7102
0.319	8.0	902	0.7763	0.7400
0.1876	8.99	1014	0.8912	0.7334
0.1888	10.0	1127	0.8836	0.7312
0.1526	11.0	1240	1.0474	0.7290
0.0451	12.0	1353	1.0455	0.7434
0.1281	12.99	1465	1.1207	0.7412
0.0363	14.0	1578	1.1232	0.7445
0.0512	14.9	1680	1.1217	0.7412

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.0
• Datasets 2.12.0
• Tokenizers 0.13.3