c14kevincardenas/deit-tiny-distilled-patch16-224_alpha0.7_temp5.0_t2

deit-tiny-distilled-patch16-224_alpha0.7_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: 0.8232
• Accuracy: 0.6789

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
1.1279	1.0	90	1.4258	0.2885
1.0009	2.0	180	1.2871	0.4032
0.8508	3.0	270	1.1393	0.5188
0.6275	4.0	360	0.9692	0.5998
0.4415	5.0	450	0.8753	0.6294
0.3063	6.0	540	0.9110	0.6344
0.2427	7.0	630	0.8515	0.6551
0.1898	8.0	720	0.8637	0.6650
0.1703	9.0	810	0.8610	0.6739
0.152	10.0	900	0.8408	0.6848
0.1409	11.0	990	0.8493	0.6690
0.1429	12.0	1080	0.8365	0.6749
0.1266	13.0	1170	0.8371	0.6789
0.127	14.0	1260	0.8402	0.6700
0.1225	15.0	1350	0.8437	0.6779
0.1199	16.0	1440	0.8451	0.6680
0.1172	17.0	1530	0.8232	0.6789
0.1107	18.0	1620	0.8329	0.6759
0.1106	19.0	1710	0.8262	0.6749
0.1063	20.0	1800	0.8272	0.6759

Framework versions

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