segformer-b0-finetuned-serra-do-cipo-tiled-final

This model is a fine-tuned version of nvidia/mit-b0 on the Wallksss/Serra_do_Cipo dataset. It achieves the following results on the evaluation set:

• Loss: 0.0742

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: 6e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• 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
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.7608	1.0	18	1.5904
1.3493	2.0	36	1.5500
1.0919	3.0	54	1.2765
1.1371	4.0	72	1.1125
0.8422	5.0	90	0.8438
0.7513	6.0	108	0.7143
0.6757	7.0	126	0.7517
0.6331	8.0	144	0.6190
0.5039	9.0	162	0.5626
0.4162	10.0	180	0.4932
0.3874	11.0	198	0.3817
0.3531	12.0	216	0.3511
0.2899	13.0	234	0.3011
0.2401	14.0	252	0.4051
0.2052	15.0	270	0.3022
0.1919	16.0	288	0.2634
0.168	17.0	306	0.2085
0.1416	18.0	324	0.2597
0.1223	19.0	342	0.1371
0.1185	20.0	360	0.2311
0.1091	21.0	378	0.1294
0.0942	22.0	396	0.1210
0.0848	23.0	414	0.1088
0.0788	24.0	432	0.1133
0.0762	25.0	450	0.0987
0.0791	26.0	468	0.0943
0.0675	27.0	486	0.0978
0.0677	28.0	504	0.1074
0.0609	29.0	522	0.0827
0.0599	30.0	540	0.1010
0.0629	31.0	558	0.0800
0.0588	32.0	576	0.0796
0.0544	33.0	594	0.0773
0.0514	34.0	612	0.0852
0.0518	35.0	630	0.0822
0.0505	36.0	648	0.0837
0.0493	37.0	666	0.0765
0.0503	38.0	684	0.0724
0.0477	39.0	702	0.0750
0.0479	40.0	720	0.0701
0.0433	41.0	738	0.0852
0.045	42.0	756	0.0639
0.0466	43.0	774	0.0694
0.0438	44.0	792	0.0747
0.0419	45.0	810	0.0781
0.043	46.0	828	0.0693
0.0438	47.0	846	0.0742
0.0453	48.0	864	0.0731
0.0409	49.0	882	0.0680
0.0413	50.0	900	0.0742

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.1