c14kevincardenas/swin-tiny-patch4-window7-224_alpha0.5_temp5.0_t2

swin-tiny-patch4-window7-224_alpha0.5_temp5.0

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: 0.5913
• Accuracy: 0.7974

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
1.2082	1.0	90	1.4079	0.2668
1.0779	2.0	180	1.2162	0.4377
0.8302	3.0	270	0.9251	0.6166
0.628	4.0	360	0.7442	0.7144
0.4817	5.0	450	0.6761	0.7540
0.378	6.0	540	0.7098	0.7559
0.3129	7.0	630	0.6408	0.7737
0.2627	8.0	720	0.6368	0.7806
0.2458	9.0	810	0.6722	0.7698
0.2303	10.0	900	0.6332	0.7806
0.2018	11.0	990	0.6203	0.7885
0.1942	12.0	1080	0.6081	0.7885
0.1831	13.0	1170	0.6041	0.7905
0.1827	14.0	1260	0.6043	0.7895
0.1825	15.0	1350	0.5913	0.7974
0.1727	16.0	1440	0.6064	0.7974
0.1701	17.0	1530	0.6057	0.8034
0.1595	18.0	1620	0.5969	0.7925
0.1616	19.0	1710	0.5923	0.7964
0.1594	20.0	1800	0.5932	0.7964

Framework versions

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