wav2vec2-base-960h-EMOPIA

This model is a fine-tuned version of facebook/wav2vec2-base-960h on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.4624
• Accuracy: 0.6620

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: 1
• eval_batch_size: 1
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2041	1.0	807	1.1999	0.3662
1.067	2.0	1614	1.4613	0.4507
1.1007	3.0	2421	1.5213	0.4085
1.1945	4.0	3228	1.6642	0.6056
1.3665	5.0	4035	2.9908	0.4366
1.4506	6.0	4842	1.9230	0.6056
1.495	7.0	5649	1.6813	0.6761
1.2605	8.0	6456	1.8937	0.6620
1.2713	9.0	7263	1.6284	0.6901
1.2608	10.0	8070	1.9438	0.6479
1.2068	11.0	8877	1.5237	0.7183
1.0478	12.0	9684	2.0007	0.6338
1.1282	13.0	10491	1.5307	0.7465
0.9433	14.0	11298	2.0042	0.6479
0.9574	15.0	12105	2.1985	0.6338
0.8737	16.0	12912	2.1568	0.6479
0.8937	17.0	13719	2.2980	0.6197
0.8681	18.0	14526	2.3268	0.6197
0.8005	19.0	15333	2.4827	0.6479
0.8176	20.0	16140	2.4842	0.6338
0.8133	21.0	16947	2.0620	0.6761
0.7404	22.0	17754	2.4148	0.6479
0.7134	23.0	18561	2.3389	0.6761
0.6573	24.0	19368	2.6972	0.6197
0.6848	25.0	20175	2.3375	0.6761
0.6161	26.0	20982	2.4791	0.6620
0.6301	27.0	21789	2.3807	0.6479
0.5758	28.0	22596	2.2243	0.6901
0.5598	29.0	23403	2.4130	0.6620
0.6066	30.0	24210	2.4624	0.6620

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu118
• Datasets 2.21.0
• Tokenizers 0.19.1