c14kevincardenas/deit-small-distilled-patch16-224_alpha0.5_temp3.0_t2

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

This model is a fine-tuned version of c14kevincardenas/ClimBEiTv2 on the c14kevincardenas/beta_caller_284_person_crop_seq_withlimb_2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6994
• Accuracy: 0.7352

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: 64
• eval_batch_size: 64
• 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
1.2256	1.0	90	1.4204	0.2708
1.0697	2.0	180	1.2418	0.4377
0.745	3.0	270	0.8560	0.6344
0.4822	4.0	360	0.7597	0.7055
0.296	5.0	450	0.7598	0.7036
0.2083	6.0	540	0.7894	0.6868
0.1911	7.0	630	0.7575	0.7134
0.1783	8.0	720	0.7299	0.7302
0.1743	9.0	810	0.7375	0.7184
0.1645	10.0	900	0.7215	0.7273
0.1548	11.0	990	0.7100	0.7411
0.1544	12.0	1080	0.7043	0.7352
0.1461	13.0	1170	0.7097	0.7352
0.1439	14.0	1260	0.7007	0.7431
0.1428	15.0	1350	0.7013	0.7401
0.1409	16.0	1440	0.7026	0.7401
0.1402	17.0	1530	0.7042	0.7431
0.1325	18.0	1620	0.6994	0.7352
0.1364	19.0	1710	0.7015	0.7460
0.1329	20.0	1800	0.7007	0.7460

Framework versions

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