diff options
Diffstat (limited to 'training/experiments')
| -rw-r--r-- | training/experiments/image_transformer.yaml | 12 |
1 files changed, 6 insertions, 6 deletions
diff --git a/training/experiments/image_transformer.yaml b/training/experiments/image_transformer.yaml index 7f0bbb7..012a19b 100644 --- a/training/experiments/image_transformer.yaml +++ b/training/experiments/image_transformer.yaml @@ -1,6 +1,6 @@ network: type: ImageTransformer - args: + args: input_shape: None output_shape: None encoder: @@ -17,20 +17,20 @@ network: model: type: LitTransformerModel args: - optimizer: + optimizer: type: MADGRAD args: lr: 1.0e-2 momentum: 0.9 weight_decay: 0 eps: 1.0e-6 - lr_scheduler: + lr_scheduler: type: CosineAnnealingLR - args: + args: T_max: 512 criterion: type: CrossEntropyLoss - args: + args: weight: None ignore_index: -100 reduction: mean @@ -40,7 +40,7 @@ model: data: type: IAMExtendedParagraphs - args: + args: batch_size: 16 num_workers: 12 train_fraction: 0.8 |