emotion_classification

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 1.1136
• Accuracy: 0.65
• F1: 0.6231

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 45
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
1.9172	1.0	43	1.5751	0.4333	0.3263
1.4505	2.0	86	1.3041	0.5333	0.4651
1.1121	3.0	129	1.2902	0.4833	0.4684
0.8491	4.0	172	1.2309	0.5167	0.4916
0.6168	5.0	215	1.2573	0.5583	0.5310
0.3953	6.0	258	1.1502	0.575	0.5401
0.3048	7.0	301	1.1136	0.65	0.6231
0.1875	8.0	344	1.4224	0.5667	0.5598
0.1277	9.0	387	1.3467	0.6167	0.6011
0.1123	10.0	430	1.5838	0.5833	0.5657
0.1123	11.0	473	1.5063	0.5833	0.5550
0.0694	12.0	516	1.7733	0.55	0.5320
0.0499	13.0	559	1.6329	0.5833	0.5536
0.0367	14.0	602	1.6878	0.5833	0.5685
0.0291	15.0	645	1.6855	0.575	0.5392
0.0284	16.0	688	1.7869	0.6083	0.5880
0.0316	17.0	731	1.5831	0.5917	0.5670
0.0273	18.0	774	1.5933	0.625	0.5984
0.0234	19.0	817	1.7830	0.5833	0.5652
0.0194	20.0	860	1.6804	0.6083	0.5878
0.0214	21.0	903	1.5962	0.6	0.5701
0.0204	22.0	946	1.5684	0.625	0.5992
0.0178	23.0	989	1.5924	0.625	0.5992
0.0173	24.0	1032	1.6228	0.6167	0.5933
0.016	25.0	1075	1.6177	0.6333	0.6073
0.016	26.0	1118	1.6268	0.625	0.6009
0.016	27.0	1161	1.6387	0.625	0.6009
0.0159	28.0	1204	1.6403	0.625	0.6009
0.0162	29.0	1247	1.6409	0.625	0.6009
0.018	30.0	1290	1.6412	0.625	0.6009

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.2
• Datasets 2.16.1
• Tokenizers 0.15.1