c14kevincardenas/deit-tiny-distilled-patch16-224_alpha0.5_temp5.0_t2

deit-tiny-distilled-patch16-224_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: 0.8025
• Accuracy: 0.6848

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
1.207	1.0	90	1.4125	0.2895
1.0715	2.0	180	1.2814	0.4121
0.9394	3.0	270	1.1279	0.5049
0.6911	4.0	360	0.9337	0.6215
0.4932	5.0	450	0.8380	0.6551
0.3481	6.0	540	0.8896	0.6433
0.2781	7.0	630	0.8509	0.6621
0.2196	8.0	720	0.8348	0.6749
0.2002	9.0	810	0.8325	0.6640
0.1821	10.0	900	0.8580	0.6700
0.1722	11.0	990	0.8279	0.6818
0.1726	12.0	1080	0.8298	0.6897
0.1583	13.0	1170	0.8128	0.6838
0.1596	14.0	1260	0.8156	0.6759
0.1551	15.0	1350	0.8226	0.6868
0.1518	16.0	1440	0.8178	0.6858
0.1485	17.0	1530	0.8045	0.6838
0.1404	18.0	1620	0.8061	0.6769
0.1415	19.0	1710	0.8038	0.6877
0.137	20.0	1800	0.8025	0.6848

Framework versions

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