finetuned-beit-limb-seq-t2-8heads-2layers-1e-4lr

This model is a fine-tuned version of c14kevincardenas/beit-large-patch16-384-limb-person-crop on the c14kevincardenas/beta_caller_284_person_crop_seq_withlimb_2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3969
• Accuracy: 0.9101

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: 0.0001
• train_batch_size: 32
• eval_batch_size: 32
• seed: 2014
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• num_epochs: 20.0
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.05

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5275	1.0	180	0.4322	0.9121
0.4514	2.0	360	0.4690	0.8962
0.425	3.0	540	0.4611	0.8903
0.4486	4.0	720	0.4024	0.9091
0.4065	5.0	900	0.4267	0.9042
0.4179	6.0	1080	0.4074	0.9061
0.4121	7.0	1260	0.4137	0.9091
0.4193	8.0	1440	0.4148	0.9130
0.3999	9.0	1620	0.4047	0.9140
0.3928	10.0	1800	0.4052	0.9170
0.3709	11.0	1980	0.4178	0.9130
0.3818	12.0	2160	0.4036	0.9170
0.3584	13.0	2340	0.4051	0.9140
0.3817	14.0	2520	0.3969	0.9101
0.3281	15.0	2700	0.4321	0.9061
0.3422	16.0	2880	0.4038	0.9140
0.3143	17.0	3060	0.4061	0.9140
0.3634	18.0	3240	0.4057	0.9101
0.3303	19.0	3420	0.4074	0.9121
0.3217	20.0	3600	0.4038	0.9130

Framework versions

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