glpn-nyu-finetuned-diode-221221-102136
This model is a fine-tuned version of vinvino02/glpn-nyu on the diode-subset dataset.
It achieves the following results on the evaluation set:
- Loss: 0.4222
- Mae: 0.4110
- Rmse: 0.6292
- Abs Rel: 0.3778
- Log Mae: 0.1636
- Log Rmse: 0.2240
- Delta1: 0.4320
- Delta2: 0.6806
- Delta3: 0.8068
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: 0.0005
- train_batch_size: 24
- eval_batch_size: 48
- seed: 2022
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.15
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Mae | Rmse | Abs Rel | Log Mae | Log Rmse | Delta1 | Delta2 | Delta3 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.4953 | 1.0 | 72 | 0.4281 | 0.4216 | 0.6448 | 0.3539 | 0.1696 | 0.2312 | 0.4427 | 0.6625 | 0.7765 |
| 0.3855 | 2.0 | 144 | 0.4749 | 0.4444 | 0.6498 | 0.4156 | 0.1846 | 0.2408 | 0.3612 | 0.6027 | 0.7728 |
| 0.4158 | 3.0 | 216 | 0.5042 | 0.5122 | 0.7196 | 0.4385 | 0.2264 | 0.2834 | 0.2797 | 0.4837 | 0.6699 |
| 0.388 | 4.0 | 288 | 0.4418 | 0.4304 | 0.6473 | 0.4030 | 0.1745 | 0.2378 | 0.4027 | 0.6497 | 0.7900 |
| 0.4595 | 5.0 | 360 | 0.4394 | 0.4154 | 0.6292 | 0.4012 | 0.1664 | 0.2285 | 0.4262 | 0.6613 | 0.8021 |
| 0.393 | 6.0 | 432 | 0.4252 | 0.4060 | 0.6153 | 0.3944 | 0.1617 | 0.2215 | 0.4318 | 0.6747 | 0.8128 |
| 0.3468 | 7.0 | 504 | 0.4413 | 0.4366 | 0.6479 | 0.3835 | 0.1818 | 0.2385 | 0.3778 | 0.6248 | 0.7770 |
| 0.316 | 8.0 | 576 | 0.4218 | 0.4048 | 0.6192 | 0.3844 | 0.1606 | 0.2215 | 0.4374 | 0.6896 | 0.8119 |
| 0.3123 | 9.0 | 648 | 0.4263 | 0.4168 | 0.6295 | 0.3765 | 0.1689 | 0.2267 | 0.4139 | 0.6612 | 0.7976 |
| 0.2973 | 10.0 | 720 | 0.4222 | 0.4110 | 0.6292 | 0.3778 | 0.1636 | 0.2240 | 0.4320 | 0.6806 | 0.8068 |
数据统计
相关导航
暂无评论...