README.md

metadata

library_name: transformers
license: other
base_model: nvidia/mit-b1
tags:
  - image-segmentation
  - vision
  - generated_from_trainer
model-index:
  - name: segformer-finetuned-tt-2k-b1
    results: []

segformer-finetuned-tt-2k-b1

This model is a fine-tuned version of nvidia/mit-b1 on the Saumya-Mundra/text255 dataset. It achieves the following results on the evaluation set:

• Loss: 0.0929
• Mean Iou: 0.4897
• Mean Accuracy: 0.9793
• Overall Accuracy: 0.9793
• Accuracy Text: nan
• Accuracy No Text: 0.9793
• Iou Text: 0.0
• Iou No Text: 0.9793

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-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 1337
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: polynomial
• training_steps: 2000

Training results

Training Loss	Epoch	Step	Accuracy No Text	Accuracy Text	Iou No Text	Iou Text	Validation Loss	Mean Accuracy	Mean Iou	Overall Accuracy
0.3305	1.0	125	0.9586	nan	0.9586	0.0	0.1846	0.9586	0.4793	0.9586
0.2037	2.0	250	0.9706	nan	0.9706	0.0	0.1322	0.9706	0.4853	0.9706
0.1534	3.0	375	0.9784	nan	0.9784	0.0	0.1074	0.9784	0.4892	0.9784
0.1313	4.0	500	0.9839	nan	0.9839	0.0	0.0976	0.9839	0.4920	0.9839
0.1156	5.0	625	0.9799	nan	0.9799	0.0	0.1001	0.9799	0.4900	0.9799
0.1123	6.0	750	0.9866	nan	0.9866	0.0	0.0920	0.9866	0.4933	0.9866
0.108	7.0	875	0.9815	nan	0.9815	0.0	0.0946	0.9815	0.4908	0.9815
0.1017	8.0	1000	0.9805	nan	0.9805	0.0	0.0943	0.9805	0.4903	0.9805
0.0994	9.0	1125	0.9808	nan	0.9808	0.0	0.0927	0.9808	0.4904	0.9808
0.0926	10.0	1250	0.9783	nan	0.9783	0.0	0.0957	0.9783	0.4891	0.9783
0.0907	11.0	1375	0.9830	nan	0.9830	0.0	0.0913	0.9830	0.4915	0.9830
0.0893	12.0	1500	0.9838	nan	0.9838	0.0	0.0893	0.9838	0.4919	0.9838
0.0853	13.0	1625	0.9804	nan	0.9804	0.0	0.0913	0.9804	0.4902	0.9804
0.0834	14.0	1750	0.9820	nan	0.9820	0.0	0.0899	0.9820	0.4910	0.9820
0.0861	15.0	1875	0.9815	nan	0.9815	0.0	0.0902	0.9815	0.4907	0.9815
0.0803	16.0	2000	0.9793	nan	0.9793	0.0	0.0929	0.9793	0.4897	0.9793

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0