swiftformer-xs-RD

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: 0.4005
• Accuracy: 0.8618

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
1.5571	0.99	40	1.5191	0.5182
0.979	1.99	80	0.8953	0.7091
0.8343	2.98	120	0.7520	0.7273
0.7755	4.0	161	0.7223	0.7527
0.7064	4.99	201	0.6760	0.74
0.6296	5.99	241	0.5624	0.8018
0.6387	6.98	281	0.5362	0.8073
0.5793	8.0	322	0.5242	0.7964
0.5723	8.99	362	0.4544	0.8309
0.5897	9.99	402	0.4510	0.8273
0.5443	10.98	442	0.4685	0.8055
0.4959	12.0	483	0.4248	0.8255
0.4743	12.99	523	0.4308	0.8309
0.4679	13.99	563	0.3999	0.84
0.4997	14.98	603	0.4151	0.8382
0.4602	16.0	644	0.4112	0.8218
0.4459	16.99	684	0.4196	0.8345
0.4668	17.99	724	0.4042	0.8291
0.41	18.98	764	0.4112	0.8436
0.4349	20.0	805	0.4120	0.8418
0.4558	20.99	845	0.4554	0.8218
0.3952	21.99	885	0.3775	0.8509
0.3395	22.98	925	0.4027	0.84
0.3525	24.0	966	0.3957	0.8509
0.3497	24.99	1006	0.4020	0.8527
0.3625	25.99	1046	0.3910	0.8564
0.3008	26.98	1086	0.3790	0.8582
0.2907	28.0	1127	0.3831	0.8564
0.3165	28.99	1167	0.4005	0.8618
0.3231	29.99	1207	0.4512	0.8327
0.2819	30.98	1247	0.4228	0.8455
0.2704	32.0	1288	0.4074	0.86
0.2429	32.99	1328	0.4405	0.8545
0.2421	33.99	1368	0.4337	0.8527
0.3039	34.98	1408	0.4628	0.8473
0.2677	36.0	1449	0.4411	0.8545
0.2171	36.99	1489	0.4755	0.8618
0.2268	37.99	1529	0.4506	0.86
0.2378	38.98	1569	0.4633	0.8527
0.2021	39.75	1600	0.4492	0.8564

Framework versions

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