diff options
Diffstat (limited to 'training/conf/network/vit_lines.yaml')
| -rw-r--r-- | training/conf/network/vit_lines.yaml | 37 |
1 files changed, 37 insertions, 0 deletions
diff --git a/training/conf/network/vit_lines.yaml b/training/conf/network/vit_lines.yaml new file mode 100644 index 0000000..35f83c3 --- /dev/null +++ b/training/conf/network/vit_lines.yaml @@ -0,0 +1,37 @@ +_target_: text_recognizer.network.vit.VisionTransformer +image_height: 56 +image_width: 1024 +patch_height: 28 +patch_width: 32 +dim: &dim 256 +num_classes: &num_classes 57 +encoder: + _target_: text_recognizer.network.transformer.encoder.Encoder + dim: *dim + inner_dim: 1024 + heads: 8 + dim_head: 64 + depth: 6 + dropout_rate: 0.0 +decoder: + _target_: text_recognizer.network.transformer.decoder.Decoder + dim: *dim + inner_dim: 1024 + heads: 8 + dim_head: 64 + depth: 6 + dropout_rate: 0.0 +token_embedding: + _target_: "text_recognizer.network.transformer.embedding.token.\ + TokenEmbedding" + num_tokens: *num_classes + dim: *dim + use_l2: true +pos_embedding: + _target_: "text_recognizer.network.transformer.embedding.absolute.\ + AbsolutePositionalEmbedding" + dim: *dim + max_length: 89 + use_l2: true +tie_embeddings: true +pad_index: 3 |