wav2vec2-xlsr-greek-speech-emotion-recognition

This model is a fine-tuned version of lighteternal/wav2vec2-large-xlsr-53-greek on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1800
• Accuracy: 0.3833

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: 0.001
• train_batch_size: 5
• eval_batch_size: 5
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 10
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 6.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.9021	0.3030	10	1.8892	0.2000
1.8885	0.6061	20	1.6402	0.2167
1.754	0.9091	30	1.6901	0.2000
1.6619	1.2121	40	1.5932	0.2167
1.652	1.5152	50	1.5049	0.3000
1.4988	1.8182	60	1.6048	0.2667
1.4969	2.1212	70	1.3770	0.3500
1.4804	2.4242	80	1.5712	0.2000
1.4504	2.7273	90	1.7242	0.2667
1.4986	3.0303	100	1.3580	0.4000
1.3599	3.3333	110	1.3808	0.2833
1.2302	3.6364	120	1.3788	0.3000
1.2667	3.9394	130	1.3989	0.3167
1.1758	4.2424	140	1.3644	0.3167
1.2662	4.5455	150	1.2385	0.3000
1.1174	4.8485	160	1.1787	0.3333
1.1285	5.1515	170	1.2276	0.3500
1.2251	5.4545	180	1.1819	0.3667
1.1332	5.7576	190	1.1800	0.3833

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1