frost-vision-v2-google_vit-base-patch16-224

This model is a fine-tuned version of google/vit-base-patch16-224 on the webdataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.1562
• Accuracy: 0.9359
• F1: 0.8381
• Precision: 0.8896
• Recall: 0.7922

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• 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
0.3416	1.1494	100	0.3273	0.8771	0.6124	0.9005	0.4640
0.2215	2.2989	200	0.2187	0.9183	0.7902	0.8537	0.7355
0.1753	3.4483	300	0.1899	0.9238	0.8098	0.8472	0.7756
0.1656	4.5977	400	0.1732	0.9272	0.8175	0.8606	0.7784
0.1288	5.7471	500	0.1562	0.9359	0.8381	0.8896	0.7922
0.1323	6.8966	600	0.1597	0.9322	0.8326	0.8609	0.8061
0.1004	8.0460	700	0.1613	0.9316	0.8324	0.8542	0.8116
0.0956	9.1954	800	0.1612	0.9336	0.8368	0.8620	0.8130
0.0841	10.3448	900	0.1621	0.9345	0.8383	0.8669	0.8116
0.0764	11.4943	1000	0.1586	0.9359	0.8438	0.8615	0.8269
0.0726	12.6437	1100	0.1546	0.9420	0.8594	0.8729	0.8463
0.0732	13.7931	1200	0.1529	0.9409	0.8565	0.87	0.8435
0.0626	14.9425	1300	0.1589	0.9377	0.8485	0.8637	0.8338
0.0481	16.0920	1400	0.1612	0.9394	0.8510	0.8767	0.8269
0.0507	17.2414	1500	0.1679	0.9339	0.8394	0.8539	0.8255
0.0446	18.3908	1600	0.1623	0.9417	0.8597	0.8664	0.8532
0.0498	19.5402	1700	0.1625	0.9417	0.8601	0.8643	0.8560
0.0458	20.6897	1800	0.1601	0.9397	0.8533	0.8693	0.8380
0.0307	21.8391	1900	0.1626	0.9432	0.8637	0.8673	0.8601
0.0334	22.9885	2000	0.1621	0.9443	0.8642	0.8829	0.8463
0.0339	24.1379	2100	0.1680	0.9435	0.8645	0.8675	0.8615
0.0222	25.2874	2200	0.1656	0.9394	0.8537	0.8628	0.8449
0.026	26.4368	2300	0.1687	0.9386	0.8515	0.8612	0.8421
0.0353	27.5862	2400	0.1666	0.9403	0.8555	0.8665	0.8449
0.0294	28.7356	2500	0.1660	0.9429	0.8614	0.8755	0.8476
0.0243	29.8851	2600	0.1664	0.9423	0.8590	0.8795	0.8393

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3