limbxy_pose_2heads_1layers_32embeddim

This model is a fine-tuned version of c14kevincardenas/beit-large-patch16-384-limb on the c14kevincardenas/beta_caller_284_limbxy_pose dataset. It achieves the following results on the evaluation set:

• Loss: 0.1391
• Rmse: 0.3729

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: 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

Training results

Training Loss	Epoch	Step	Validation Loss	Rmse
0.214	1.0	89	0.1933	0.4396
0.161	2.0	178	0.1474	0.3839
0.1634	3.0	267	0.1541	0.3925
0.1547	4.0	356	0.1495	0.3867
0.1674	5.0	445	0.1680	0.4098
0.1644	6.0	534	0.1473	0.3838
0.1459	7.0	623	0.1405	0.3748
0.1544	8.0	712	0.1514	0.3891
0.1531	9.0	801	0.1419	0.3767
0.1468	10.0	890	0.1422	0.3770
0.143	11.0	979	0.1447	0.3804
0.1467	12.0	1068	0.1597	0.3996
0.149	13.0	1157	0.1394	0.3733
0.139	14.0	1246	0.1412	0.3758
0.1433	15.0	1335	0.1397	0.3738
0.1426	16.0	1424	0.1396	0.3736
0.1403	17.0	1513	0.1396	0.3737
0.1393	18.0	1602	0.1393	0.3733
0.1426	19.0	1691	0.1391	0.3730
0.1387	20.0	1780	0.1391	0.3729

Framework versions

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