vit-base-patch32-224-in21-leicester_binary

This model is a fine-tuned version of google/vit-base-patch32-224-in21k on the davanstrien/leicester_loaded_annotations_binary dataset. It achieves the following results on the evaluation set:

• Loss: 0.0628
• F1: 0.9873

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: 2e-05
• train_batch_size: 64
• eval_batch_size: 128
• seed: 1337
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 40.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1
No log	1.0	7	0.4529	0.8608
0.5024	2.0	14	0.3989	0.8608
0.3533	3.0	21	0.3741	0.8608
0.3533	4.0	28	0.3161	0.8608
0.285	5.0	35	0.2824	0.8608
0.2491	6.0	42	0.2701	0.8608
0.2491	7.0	49	0.2062	0.9114
0.2032	8.0	56	0.2050	0.9494
0.157	9.0	63	0.2013	0.9494
0.1127	10.0	70	0.1960	0.9367
0.1127	11.0	77	0.1417	0.9494
0.0903	12.0	84	0.1307	0.9494
0.0922	13.0	91	0.0870	0.9873
0.0922	14.0	98	0.2048	0.9241
0.0595	15.0	105	0.1204	0.9620
0.0527	16.0	112	0.2553	0.9367
0.0527	17.0	119	0.1675	0.9367
0.0477	18.0	126	0.2265	0.9241
0.0411	19.0	133	0.1901	0.9367
0.0299	20.0	140	0.2423	0.9241
0.0299	21.0	147	0.0639	0.9873
0.0487	22.0	154	0.1255	0.9494
0.0359	23.0	161	0.1213	0.9494
0.0359	24.0	168	0.0727	0.9747
0.0302	25.0	175	0.1116	0.9494
0.0304	26.0	182	0.1062	0.9494
0.0304	27.0	189	0.2097	0.9241
0.0274	28.0	196	0.1276	0.9494
0.0291	29.0	203	0.0967	0.9494
0.0202	30.0	210	0.0765	0.9747
0.0202	31.0	217	0.0628	0.9873
0.0232	32.0	224	0.1388	0.9494
0.0264	33.0	231	0.1062	0.9494
0.0264	34.0	238	0.1320	0.9494
0.0219	35.0	245	0.1528	0.9494
0.0194	36.0	252	0.1746	0.9494
0.0194	37.0	259	0.1609	0.9494
0.0204	38.0	266	0.1482	0.9494
0.0217	39.0	273	0.1522	0.9494
0.0216	40.0	280	0.1499	0.9494

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2