wav2vec2-base-960h-EMOPIA-10sec

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: 2.5866
• Accuracy: 0.6338

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
1.2014	1.0	807	1.1830	0.3662
1.0915	2.0	1614	1.5120	0.3239
1.1433	3.0	2421	1.5699	0.4085
1.2819	4.0	3228	1.7372	0.4789
1.2718	5.0	4035	2.2169	0.4648
1.4535	6.0	4842	1.7296	0.5775
1.3433	7.0	5649	2.2684	0.5493
1.4086	8.0	6456	1.8599	0.6479
1.3923	9.0	7263	1.9420	0.6197
1.3353	10.0	8070	2.2150	0.5775
1.367	11.0	8877	1.9826	0.6338
1.1848	12.0	9684	1.9545	0.6479
1.1355	13.0	10491	1.9864	0.6620
1.1549	14.0	11298	1.9428	0.6338
1.0505	15.0	12105	1.9101	0.6901
1.0442	16.0	12912	2.1706	0.6479
0.9922	17.0	13719	2.4620	0.6197
0.8698	18.0	14526	2.1429	0.6620
0.8202	19.0	15333	2.3725	0.6197
0.8612	20.0	16140	2.1631	0.6620
0.8197	21.0	16947	2.3932	0.6338
0.7858	22.0	17754	2.2532	0.6479
0.7717	23.0	18561	2.8132	0.5634
0.6282	24.0	19368	2.5493	0.6197
0.7394	25.0	20175	2.3195	0.6620
0.5895	26.0	20982	2.4331	0.6620
0.5854	27.0	21789	2.4281	0.6761
0.6911	28.0	22596	2.4993	0.6620
0.5502	29.0	23403	2.6458	0.6338
0.584	30.0	24210	2.5866	0.6338

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu118
• Datasets 3.0.1
• Tokenizers 0.20.0