mit-b0-finetuned-sidewalk-semantic

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

• Train Loss: 0.2125
• Validation Loss: 0.5151
• Epoch: 49

Model description

The model was fine-tuned from this model. More information about the model is available here.

Intended uses & limitations

This fine-tuned model is just for demonstration purposes. Before using it in production, it should be thoroughly inspected and adjusted if needed.

Training and evaluation data

segments/sidewalk-semantic

Training procedure

More information is available here: deep-diver/segformer-tf-transformers.

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'Adam', 'learning_rate': 6e-05, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
• training_precision: float32

Training results

Train Loss	Validation Loss	Epoch
2.0785	1.1753	0
1.1312	0.8807	1
0.9315	0.7585	2
0.7952	0.7261	3
0.7273	0.6701	4
0.6603	0.6396	5
0.6198	0.6238	6
0.5958	0.5925	7
0.5378	0.5714	8
0.5236	0.5786	9
0.4960	0.5588	10
0.4633	0.5624	11
0.4562	0.5450	12
0.4167	0.5438	13
0.4100	0.5248	14
0.3947	0.5354	15
0.3867	0.5069	16
0.3803	0.5285	17
0.3696	0.5318	18
0.3386	0.5162	19
0.3349	0.5312	20
0.3233	0.5304	21
0.3328	0.5178	22
0.3140	0.5131	23
0.3081	0.5049	24
0.3046	0.5011	25
0.3209	0.5197	26
0.2966	0.5151	27
0.2829	0.5166	28
0.2968	0.5210	29
0.2818	0.5300	30
0.2739	0.5221	31
0.2602	0.5340	32
0.2570	0.5124	33
0.2557	0.5234	34
0.2593	0.5098	35
0.2582	0.5329	36
0.2439	0.5373	37
0.2413	0.5141	38
0.2423	0.5210	39
0.2340	0.5043	40
0.2244	0.5300	41
0.2246	0.4978	42
0.2270	0.5385	43
0.2254	0.5125	44
0.2176	0.5510	45
0.2194	0.5384	46
0.2136	0.5186	47
0.2121	0.5356	48
0.2125	0.5151	49

Framework versions

• Transformers 4.21.0.dev0
• TensorFlow 2.8.0
• Datasets 2.3.2
• Tokenizers 0.12.1