c14kevincardenas/mobilevit-small_alpha0.7_temp3.0_t2

mobilevit-small_alpha0.7_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: 0.8524
• Accuracy: 0.6660

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
1.1499	1.0	90	1.4251	0.2460
1.0997	2.0	180	1.3384	0.3251
1.0075	3.0	270	1.2409	0.3864
0.8689	4.0	360	1.0919	0.4980
0.7581	5.0	450	1.0005	0.5652
0.7002	6.0	540	0.9315	0.6038
0.6484	7.0	630	0.9005	0.6166
0.6085	8.0	720	0.9252	0.5988
0.5562	9.0	810	0.8605	0.6630
0.5352	10.0	900	0.8696	0.6265
0.514	11.0	990	0.8557	0.6571
0.4819	12.0	1080	0.8793	0.6294
0.4538	13.0	1170	0.8856	0.6354
0.4382	14.0	1260	0.8554	0.6581
0.4349	15.0	1350	0.8707	0.6393
0.4122	16.0	1440	0.8568	0.6640
0.4101	17.0	1530	0.8540	0.6630
0.3922	18.0	1620	0.8543	0.6621
0.3929	19.0	1710	0.8524	0.6660
0.3965	20.0	1800	0.8661	0.6512

Framework versions

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