deit-small-distilled-patch16-224_alpha0.5_temp3.0_t3

This model is a fine-tuned version of c14kevincardenas/ClimBEiT-t3 on the c14kevincardenas/beta_caller_284_person_crop_seq_withlimb dataset. It achieves the following results on the evaluation set:

• Loss: 0.6687
• Accuracy: 0.8113

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.8112	1.0	164	1.4054	0.2505
0.7489	2.0	328	1.2976	0.3601
0.5759	3.0	492	0.8688	0.7191
0.4444	4.0	656	0.7477	0.7614
0.3614	5.0	820	0.7505	0.7690
0.3219	6.0	984	0.7726	0.7538
0.3063	7.0	1148	0.7297	0.7722
0.301	8.0	1312	0.7008	0.7755
0.2967	9.0	1476	0.6961	0.7842
0.2823	10.0	1640	0.6866	0.7863
0.2831	11.0	1804	0.6905	0.7918
0.2874	12.0	1968	0.6724	0.8004
0.2769	13.0	2132	0.6792	0.8048
0.2821	14.0	2296	0.6699	0.8069
0.2854	15.0	2460	0.6829	0.8069
0.2817	16.0	2624	0.6700	0.8102
0.2852	17.0	2788	0.6715	0.8102
0.2751	18.0	2952	0.6703	0.8124
0.2821	19.0	3116	0.6687	0.8113
0.2749	20.0	3280	0.6695	0.8102

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1