finetuned-beit-limb-seq-t2-4heads-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: 0.3945
• Accuracy: 0.9160

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
0.5237	1.0	180	0.4331	0.9111
0.4502	2.0	360	0.4453	0.9012
0.4249	3.0	540	0.4701	0.8903
0.4471	4.0	720	0.4051	0.9101
0.4084	5.0	900	0.4256	0.9101
0.4206	6.0	1080	0.4056	0.9091
0.407	7.0	1260	0.4148	0.9101
0.414	8.0	1440	0.4190	0.9121
0.3938	9.0	1620	0.4035	0.9130
0.3901	10.0	1800	0.4041	0.9150
0.3721	11.0	1980	0.4215	0.9130
0.3819	12.0	2160	0.4014	0.9140
0.3622	13.0	2340	0.4003	0.9160
0.3805	14.0	2520	0.3945	0.9160
0.3243	15.0	2700	0.4261	0.9081
0.3426	16.0	2880	0.3970	0.9140
0.3131	17.0	3060	0.4001	0.9180
0.3645	18.0	3240	0.4001	0.9140
0.3254	19.0	3420	0.4021	0.9111
0.3242	20.0	3600	0.3978	0.9180

Framework versions

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