glpn-nyu-finetuned-diode-230530-204740

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: 1.5139
  • Mae: 3.0509
  • Rmse: 3.4756
  • Abs Rel: 5.7613
  • Log Mae: 0.6836
  • Log Rmse: 0.8048
  • Delta1: 0.3028
  • Delta2: 0.3079
  • Delta3: 0.3096

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: 1e-05
  • 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: 0.1
  • 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
No log 1.0 1 1.5335 3.1427 3.6089 5.9847 0.6920 0.8173 0.3016 0.3077 0.3094
No log 2.0 2 1.5297 3.1246 3.5833 5.9419 0.6903 0.8149 0.3018 0.3077 0.3094
No log 3.0 3 1.5263 3.1085 3.5602 5.9033 0.6889 0.8128 0.3020 0.3078 0.3095
No log 4.0 4 1.5234 3.0947 3.5400 5.8694 0.6876 0.8109 0.3022 0.3078 0.3095
No log 5.0 5 1.5208 3.0825 3.5222 5.8395 0.6865 0.8092 0.3024 0.3079 0.3095
No log 6.0 6 1.5185 3.0723 3.5072 5.8144 0.6856 0.8078 0.3025 0.3079 0.3095
No log 7.0 7 1.5167 3.0639 3.4949 5.7937 0.6848 0.8067 0.3026 0.3079 0.3096
No log 8.0 8 1.5153 3.0574 3.4852 5.7775 0.6842 0.8057 0.3027 0.3079 0.3096
No log 9.0 9 1.5143 3.0531 3.4788 5.7667 0.6838 0.8051 0.3028 0.3079 0.3096
No log 10.0 10 1.5139 3.0509 3.4756 5.7613 0.6836 0.8048 0.3028 0.3079 0.3096

Framework versions

  • Transformers 4.29.2
  • Pytorch 2.0.1+cu118
  • Tokenizers 0.13.3
Downloads last month
12
Inference API
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.