BEiT-TO-DA

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

• Loss: 0.3226
• Accuracy: 0.9032

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.5028	0.97	14	1.3862	0.1452
1.3358	2.0	29	1.0831	0.8065
0.8919	2.97	43	0.8097	0.8226
0.7328	4.0	58	0.6546	0.7742
0.4881	4.97	72	0.4860	0.8226
0.4367	6.0	87	0.4770	0.8548
0.3343	6.97	101	0.3912	0.8548
0.2794	8.0	116	0.3226	0.9032
0.2424	8.97	130	0.6426	0.7903
0.2875	10.0	145	0.4604	0.8710
0.226	10.97	159	0.3026	0.8548
0.1819	12.0	174	0.3875	0.8710
0.2354	12.97	188	0.3413	0.9032
0.2264	14.0	203	0.3948	0.8871
0.1652	14.97	217	0.3650	0.8710
0.1449	16.0	232	0.3611	0.8871
0.0993	16.97	246	0.4574	0.8710
0.1566	18.0	261	0.3924	0.8871
0.1399	18.97	275	0.4828	0.8548
0.1025	20.0	290	0.5377	0.8710
0.0855	20.97	304	0.4958	0.8548
0.1419	22.0	319	0.6156	0.8387
0.117	22.97	333	0.4915	0.8710
0.0905	24.0	348	0.5897	0.8710
0.1199	24.97	362	0.4871	0.8710
0.1246	26.0	377	0.4824	0.8548
0.0967	26.97	391	0.7484	0.8065
0.1025	28.0	406	0.6974	0.8387
0.1112	28.97	420	0.6391	0.8226
0.0715	30.0	435	0.6585	0.8226
0.085	30.97	449	0.7087	0.8065
0.1032	32.0	464	0.6094	0.8387
0.0836	32.97	478	0.5578	0.8065
0.0716	34.0	493	0.5497	0.8710
0.069	34.97	507	0.5093	0.8710
0.0577	36.0	522	0.5189	0.8710
0.0882	36.97	536	0.6531	0.8226
0.0563	38.0	551	0.6661	0.8226
0.0841	38.62	560	0.6616	0.8226

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0