frost-mobile-apple__mobilevit-xx-small-v2024-10-22

This model is a fine-tuned version of apple/mobilevit-xx-small on the webdataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.1381
• Accuracy: 0.9476
• F1: 0.8700
• Precision: 0.8681
• Recall: 0.8720

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 8
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.1927	1.7544	100	0.1470	0.9422	0.8565	0.8565	0.8565
0.1601	3.5088	200	0.1499	0.9444	0.8616	0.8644	0.8587
0.1544	5.2632	300	0.1536	0.9391	0.8493	0.8465	0.8521
0.207	7.0175	400	0.1374	0.9436	0.8575	0.8721	0.8433
0.1709	8.7719	500	0.1443	0.9431	0.8587	0.8587	0.8587
0.1548	10.5263	600	0.1572	0.9387	0.8490	0.8416	0.8565
0.1802	12.2807	700	0.1436	0.9458	0.8656	0.8637	0.8675
0.1455	14.0351	800	0.1442	0.9467	0.8667	0.8725	0.8609
0.1514	15.7895	900	0.1500	0.9422	0.8571	0.8534	0.8609
0.1368	17.5439	1000	0.1391	0.9489	0.8718	0.8806	0.8631
0.1515	19.2982	1100	0.1370	0.9476	0.8700	0.8681	0.8720
0.1372	21.0526	1200	0.1393	0.9458	0.8644	0.8702	0.8587
0.1397	22.8070	1300	0.1359	0.9498	0.8746	0.8795	0.8698
0.1398	24.5614	1400	0.1352	0.9489	0.8740	0.8674	0.8808
0.1276	26.3158	1500	0.1381	0.9476	0.8700	0.8681	0.8720
0.1519	28.0702	1600	0.1380	0.9462	0.8666	0.8656	0.8675
0.1479	29.8246	1700	0.1343	0.9498	0.8754	0.8744	0.8764

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.2
• Tokenizers 0.19.1