beit-base-patch16-224-pt22k-ft22k-rim_one-new

This model is a fine-tuned version of microsoft/beit-base-patch16-224-pt22k-ft22k on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.4550
• Accuracy: 0.8767

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.73	2	0.2411	0.9178
No log	1.73	4	0.2182	0.8973
No log	2.73	6	0.3085	0.8973
No log	3.73	8	0.2794	0.8973
0.1392	4.73	10	0.2398	0.9110
0.1392	5.73	12	0.2925	0.8973
0.1392	6.73	14	0.2798	0.9110
0.1392	7.73	16	0.2184	0.9178
0.1392	8.73	18	0.3007	0.9110
0.0416	9.73	20	0.3344	0.9041
0.0416	10.73	22	0.3626	0.9110
0.0416	11.73	24	0.4842	0.8904
0.0416	12.73	26	0.3664	0.8973
0.0416	13.73	28	0.3458	0.9110
0.0263	14.73	30	0.2810	0.9110
0.0263	15.73	32	0.4695	0.8699
0.0263	16.73	34	0.3723	0.9041
0.0263	17.73	36	0.3447	0.9041
0.0263	18.73	38	0.3708	0.8904
0.0264	19.73	40	0.4052	0.9110
0.0264	20.73	42	0.4492	0.9041
0.0264	21.73	44	0.4649	0.8904
0.0264	22.73	46	0.4061	0.9178
0.0264	23.73	48	0.4136	0.9110
0.0139	24.73	50	0.4183	0.8973
0.0139	25.73	52	0.4504	0.8904
0.0139	26.73	54	0.4368	0.8973
0.0139	27.73	56	0.4711	0.9110
0.0139	28.73	58	0.3928	0.9110
0.005	29.73	60	0.4550	0.8767

Framework versions

• Transformers 4.20.1
• Pytorch 1.11.0+cu113
• Datasets 2.3.2
• Tokenizers 0.12.1