beit-base-finetuned-ade-640-640_alpha0.5_temp5.0_t2

This model is a fine-tuned version of c14kevincardenas/ClimBEiT-t2 on the c14kevincardenas/beta_caller_284_person_crop_seq_withlimb dataset. It achieves the following results on the evaluation set:

• Loss: 0.4849
• Accuracy: 0.8389

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: 32
• eval_batch_size: 32
• 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
1.1929	1.0	180	1.3065	0.3972
0.7186	2.0	360	0.8221	0.6917
0.5468	3.0	540	0.7486	0.7223
0.4245	4.0	720	0.6552	0.7688
0.3503	5.0	900	0.5798	0.7945
0.2533	6.0	1080	0.6044	0.7875
0.2419	7.0	1260	0.5754	0.8014
0.238	8.0	1440	0.6908	0.7727
0.2325	9.0	1620	0.5579	0.8063
0.2118	10.0	1800	0.5279	0.8261
0.1827	11.0	1980	0.5354	0.8182
0.1797	12.0	2160	0.5396	0.8310
0.1628	13.0	2340	0.5208	0.8350
0.1628	14.0	2520	0.5148	0.8182
0.1613	15.0	2700	0.5173	0.8162
0.151	16.0	2880	0.4949	0.8320
0.1492	17.0	3060	0.5020	0.8281
0.1538	18.0	3240	0.4889	0.8281
0.1434	19.0	3420	0.4910	0.8350
0.1391	20.0	3600	0.4849	0.8389

Framework versions

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