c14kevincardenas/deit-tiny-distilled-patch16-224_alpha0.7_temp3.0_t2

deit-tiny-distilled-patch16-224_alpha0.7_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: 0.8283
• Accuracy: 0.6808

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
1.1465	1.0	90	1.4257	0.2856
1.0136	2.0	180	1.2902	0.4061
0.8581	3.0	270	1.1425	0.5099
0.6313	4.0	360	0.9627	0.6008
0.4469	5.0	450	0.8729	0.6265
0.307	6.0	540	0.9136	0.6275
0.2482	7.0	630	0.8533	0.6502
0.1911	8.0	720	0.8625	0.6561
0.1729	9.0	810	0.8415	0.6759
0.1548	10.0	900	0.8361	0.6700
0.1445	11.0	990	0.8495	0.6769
0.1438	12.0	1080	0.8442	0.6680
0.1294	13.0	1170	0.8405	0.6739
0.1279	14.0	1260	0.8366	0.6779
0.1238	15.0	1350	0.8354	0.6739
0.1207	16.0	1440	0.8447	0.6630
0.1185	17.0	1530	0.8283	0.6808
0.1118	18.0	1620	0.8335	0.6670
0.1116	19.0	1710	0.8291	0.6729
0.1074	20.0	1800	0.8299	0.6719

Framework versions

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