vit-base-patch32-384-finetuned-humid-classes-26

This model is a fine-tuned version of google/vit-base-patch32-384 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.1926
• Accuracy: 0.9706
• F1 Macro: 0.9429
• Precision Macro: 0.9583
• Recall Macro: 0.9444
• Precision Dry: 0.75
• Recall Dry: 1.0
• F1 Dry: 0.8571
• Precision Firm: 1.0
• Recall Firm: 1.0
• F1 Firm: 1.0
• Precision Humid: 1.0
• Recall Humid: 1.0
• F1 Humid: 1.0
• Precision Lump: 1.0
• Recall Lump: 1.0
• F1 Lump: 1.0
• Precision Moist: 1.0
• Recall Moist: 1.0
• F1 Moist: 1.0
• Precision Rockies: 1.0
• Recall Rockies: 0.6667
• F1 Rockies: 0.8

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: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	Precision Macro	Recall Macro	Precision Dry	Recall Dry	F1 Dry	Precision Firm	Recall Firm	F1 Firm	Precision Humid	Recall Humid	F1 Humid	Precision Lump	Recall Lump	F1 Lump	Precision Moist	Recall Moist	F1 Moist	Precision Rockies	Recall Rockies	F1 Rockies
No log	1.0	3	1.8405	0.2647	0.2000	0.1850	0.3214	0.0	0.0	0.0	0.0	0.0	0.0	0.6	0.4286	0.5	0.3333	0.5	0.4	0.0	0.0	0.0	0.1765	1.0	0.3
No log	2.0	6	1.4673	0.5588	0.4141	0.4078	0.4798	0.0	0.0	0.0	1.0	0.5455	0.7059	0.5833	1.0	0.7368	0.3636	0.6667	0.4706	0.0	0.0	0.0	0.5	0.6667	0.5714
No log	3.0	9	1.1051	0.6471	0.4789	0.5540	0.5000	1.0	0.3333	0.5	0.7857	1.0	0.88	0.5385	1.0	0.7	0.3333	0.1667	0.2222	0.6667	0.5	0.5714	0.0	0.0	0.0
1.6213	4.0	12	0.7794	0.8824	0.8540	0.9077	0.8472	0.6	1.0	0.75	0.8462	1.0	0.9167	1.0	1.0	1.0	1.0	0.6667	0.8	1.0	0.75	0.8571	1.0	0.6667	0.8
1.6213	5.0	15	0.5343	0.9118	0.8858	0.9410	0.8611	1.0	0.6667	0.8	0.8462	1.0	0.9167	1.0	1.0	1.0	1.0	0.8333	0.9091	0.8	1.0	0.8889	1.0	0.6667	0.8
1.6213	6.0	18	0.3642	0.8824	0.8623	0.9083	0.8472	1.0	1.0	1.0	1.0	1.0	1.0	0.7	1.0	0.8235	1.0	0.6667	0.8	0.75	0.75	0.75	1.0	0.6667	0.8
0.6346	7.0	21	0.2741	0.9412	0.9219	0.9458	0.9167	1.0	1.0	1.0	1.0	1.0	1.0	0.875	1.0	0.9333	1.0	0.8333	0.9091	0.8	1.0	0.8889	1.0	0.6667	0.8
0.6346	8.0	24	0.1926	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.6346	9.0	27	0.1662	0.9412	0.9258	0.9528	0.9167	1.0	1.0	1.0	0.9167	1.0	0.9565	1.0	1.0	1.0	1.0	0.8333	0.9091	0.8	1.0	0.8889	1.0	0.6667	0.8
0.1013	10.0	30	0.1648	0.9412	0.9205	0.9444	0.9167	0.75	1.0	0.8571	0.9167	1.0	0.9565	1.0	1.0	1.0	1.0	0.8333	0.9091	1.0	1.0	1.0	1.0	0.6667	0.8
0.1013	11.0	33	0.0794	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.1013	12.0	36	0.0757	0.9706	0.9481	0.9667	0.9444	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.8	1.0	0.8889	1.0	0.6667	0.8
0.1013	13.0	39	0.1002	0.9706	0.9481	0.9667	0.9444	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.8	1.0	0.8889	1.0	0.6667	0.8
0.0169	14.0	42	0.1836	0.9118	0.8887	0.9213	0.8889	0.75	1.0	0.8571	1.0	1.0	1.0	0.7778	1.0	0.875	1.0	0.6667	0.8	1.0	1.0	1.0	1.0	0.6667	0.8
0.0169	15.0	45	0.2520	0.9118	0.8887	0.9213	0.8889	0.75	1.0	0.8571	1.0	1.0	1.0	0.7778	1.0	0.875	1.0	0.6667	0.8	1.0	1.0	1.0	1.0	0.6667	0.8
0.0169	16.0	48	0.1942	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0055	17.0	51	0.1598	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0055	18.0	54	0.1251	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0055	19.0	57	0.1000	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0027	20.0	60	0.0868	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0027	21.0	63	0.0766	0.9706	0.9481	0.9667	0.9444	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.8	1.0	0.8889	1.0	0.6667	0.8
0.0027	22.0	66	0.0703	0.9706	0.9481	0.9667	0.9444	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.8	1.0	0.8889	1.0	0.6667	0.8
0.0027	23.0	69	0.0701	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0022	24.0	72	0.0768	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0022	25.0	75	0.0882	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0022	26.0	78	0.0992	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0014	27.0	81	0.1085	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0014	28.0	84	0.1154	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0014	29.0	87	0.1197	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0011	30.0	90	0.1221	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0011	31.0	93	0.1232	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0011	32.0	96	0.1231	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0011	33.0	99	0.1223	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0009	34.0	102	0.1212	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0009	35.0	105	0.1204	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0009	36.0	108	0.1201	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0009	37.0	111	0.1199	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0009	38.0	114	0.1195	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0009	39.0	117	0.1190	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	40.0	120	0.1183	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	41.0	123	0.1177	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	42.0	126	0.1172	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	43.0	129	0.1167	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	44.0	132	0.1162	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	45.0	135	0.1160	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	46.0	138	0.1160	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0007	47.0	141	0.1160	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0007	48.0	144	0.1160	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0007	49.0	147	0.1160	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8
0.0008	50.0	150	0.1160	0.9706	0.9429	0.9583	0.9444	0.75	1.0	0.8571	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	1.0	0.6667	0.8

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.0