ViT-Emotion-Classifier

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: 1.3652
• Accuracy: 0.575

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	40	1.8992	0.3312
No log	2.0	80	1.5939	0.4062
No log	3.0	120	1.4776	0.4688
No log	4.0	160	1.4012	0.4813
No log	5.0	200	1.3471	0.4875
No log	6.0	240	1.2877	0.5375
No log	7.0	280	1.2598	0.575
No log	8.0	320	1.3595	0.4938
No log	9.0	360	1.2825	0.5375
No log	10.0	400	1.3291	0.5062
No log	11.0	440	1.2422	0.5563
No log	12.0	480	1.2659	0.575
1.0646	13.0	520	1.3048	0.5062
1.0646	14.0	560	1.2993	0.5563
1.0646	15.0	600	1.2935	0.5563
1.0646	16.0	640	1.3589	0.5437
1.0646	17.0	680	1.2447	0.5938
1.0646	18.0	720	1.3298	0.5563
1.0646	19.0	760	1.2829	0.6
1.0646	20.0	800	1.3092	0.5813
1.0646	21.0	840	1.2895	0.5875
1.0646	22.0	880	1.3810	0.5625
1.0646	23.0	920	1.3833	0.5563
1.0646	24.0	960	1.4841	0.5312
0.3074	25.0	1000	1.3619	0.6062
0.3074	26.0	1040	1.3776	0.5563
0.3074	27.0	1080	1.3917	0.5875
0.3074	28.0	1120	1.3585	0.575
0.3074	29.0	1160	1.3455	0.5625
0.3074	30.0	1200	1.4409	0.5813

Framework versions

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