swiftformer-xs-RH

This model is a fine-tuned version of MBZUAI/swiftformer-xs on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.4150
• Accuracy: 0.8598

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	8	0.6926	0.6262
0.693	2.0	16	0.6873	0.6822
0.6899	3.0	24	0.6694	0.6729
0.6732	4.0	32	0.6146	0.7290
0.6256	5.0	40	0.7084	0.4953
0.6256	6.0	48	0.6591	0.6355
0.5796	7.0	56	0.5670	0.7383
0.5253	8.0	64	0.5351	0.7196
0.4713	9.0	72	0.4614	0.8411
0.441	10.0	80	0.5826	0.7570
0.441	11.0	88	0.4679	0.7850
0.3803	12.0	96	0.4517	0.8411
0.3513	13.0	104	0.4571	0.7757
0.3043	14.0	112	0.4755	0.8037
0.3172	15.0	120	0.5953	0.7944
0.3172	16.0	128	0.5904	0.7383
0.365	17.0	136	0.4213	0.8411
0.279	18.0	144	0.4572	0.8037
0.3092	19.0	152	0.4181	0.8131
0.2667	20.0	160	0.4117	0.8224
0.2667	21.0	168	0.4349	0.8037
0.261	22.0	176	0.4185	0.8037
0.2638	23.0	184	0.3989	0.8131
0.2269	24.0	192	0.3971	0.8318
0.2431	25.0	200	0.4784	0.8037
0.2431	26.0	208	0.3763	0.8318
0.2111	27.0	216	0.4088	0.8411
0.2087	28.0	224	0.4024	0.8318
0.1645	29.0	232	0.4161	0.8318
0.2137	30.0	240	0.4128	0.8131
0.2137	31.0	248	0.4004	0.8411
0.2207	32.0	256	0.4206	0.8224
0.1544	33.0	264	0.3622	0.8505
0.197	34.0	272	0.4356	0.8411
0.2168	35.0	280	0.4067	0.8318
0.2168	36.0	288	0.3809	0.8224
0.1631	37.0	296	0.3865	0.8411
0.1913	38.0	304	0.4008	0.8318
0.1595	39.0	312	0.3752	0.8318
0.1694	40.0	320	0.4150	0.8598

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0