emotion_image_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: 1.3343
• Accuracy: 0.5875

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: 0.0005
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 5
• total_train_batch_size: 160
• 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	0.87	4	2.0221	0.1
No log	1.96	9	1.6982	0.25
No log	2.83	13	1.7868	0.225
No log	3.91	18	1.6731	0.2625
No log	5.0	23	1.6196	0.175
No log	5.87	27	1.5399	0.3
No log	6.96	32	1.5348	0.375
No log	7.83	36	1.6157	0.3125
No log	8.91	41	1.4275	0.45
No log	10.0	46	1.3832	0.425
No log	10.87	50	1.4440	0.425
No log	11.96	55	1.5841	0.4375
No log	12.83	59	1.4398	0.4625
No log	13.91	64	1.4413	0.475
No log	15.0	69	1.3143	0.5375
No log	15.87	73	1.3667	0.5625
No log	16.96	78	1.4028	0.5
No log	17.83	82	1.4485	0.5375
No log	18.91	87	1.9334	0.3875
No log	20.0	92	1.4611	0.55
No log	20.87	96	1.3279	0.5875
No log	21.96	101	1.6526	0.45
No log	22.83	105	1.4921	0.4875
No log	23.91	110	1.3962	0.5875
No log	25.0	115	1.7038	0.4375
No log	25.87	119	1.5210	0.55
No log	26.09	120	1.5141	0.5125

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.1