BEiT-RHS-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: 1.7194
• Accuracy: 0.6355

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
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2357	0.98	22	0.7114	0.5888
0.6596	2.0	45	0.7059	0.5981
0.206	2.98	67	1.1449	0.5981
0.1664	4.0	90	2.2062	0.3925
0.1011	4.98	112	2.0409	0.4673
0.0653	6.0	135	1.3038	0.6262
0.2843	6.98	157	1.7210	0.5981
0.059	8.0	180	2.8706	0.4673
0.1318	8.98	202	2.4519	0.5888
0.0501	10.0	225	2.2037	0.5888
0.054	10.98	247	2.6467	0.5888
0.0263	12.0	270	2.4033	0.5981
0.0553	12.98	292	1.6589	0.5888
0.0898	14.0	315	1.7657	0.5981
0.0324	14.98	337	2.8266	0.5888
0.0322	16.0	360	1.7194	0.6355
0.03	16.98	382	2.0352	0.6168
0.0392	18.0	405	2.4130	0.6168
0.0428	18.98	427	2.0628	0.6075
0.0127	20.0	450	2.7431	0.5888
0.0187	20.98	472	2.7009	0.5981
0.0469	22.0	495	2.5783	0.5981
0.0095	22.98	517	2.3040	0.5981
0.0025	24.0	540	2.5218	0.6168
0.0281	24.98	562	3.2310	0.5981
0.0004	26.0	585	3.2731	0.5981
0.0109	26.98	607	2.4809	0.6262
0.0191	28.0	630	2.7825	0.5888
0.0005	28.98	652	3.5280	0.5888
0.0093	30.0	675	2.8290	0.6075
0.0224	30.98	697	2.9546	0.5794
0.0011	32.0	720	3.0148	0.6075
0.003	32.98	742	3.2916	0.5981
0.0003	34.0	765	3.2930	0.5981
0.0003	34.98	787	3.6287	0.5888
0.0002	36.0	810	3.6918	0.5888
0.0004	36.98	832	3.6597	0.5888
0.0003	38.0	855	3.6599	0.5888
0.0002	38.98	877	3.6740	0.5888
0.0002	39.11	880	3.6741	0.5888

Framework versions

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