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

deit-tiny-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: 0.8034
• Accuracy: 0.6798

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
1.2203	1.0	90	1.4123	0.2875
1.0812	2.0	180	1.2838	0.4061
0.9486	3.0	270	1.1327	0.5040
0.6955	4.0	360	0.9298	0.6166
0.4946	5.0	450	0.8451	0.6551
0.3529	6.0	540	0.8921	0.6472
0.2815	7.0	630	0.8604	0.6581
0.2195	8.0	720	0.8425	0.6571
0.201	9.0	810	0.8144	0.6808
0.186	10.0	900	0.8184	0.6897
0.1747	11.0	990	0.8467	0.6709
0.1766	12.0	1080	0.8266	0.6789
0.1615	13.0	1170	0.8106	0.6729
0.1618	14.0	1260	0.8074	0.6868
0.1574	15.0	1350	0.8163	0.6789
0.1534	16.0	1440	0.8144	0.6798
0.1507	17.0	1530	0.8034	0.6798
0.1426	18.0	1620	0.8038	0.6828
0.1436	19.0	1710	0.8051	0.6848
0.1393	20.0	1800	0.8045	0.6877

Framework versions

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