summaryrefslogtreecommitdiff
path: root/training/conf/network/conformer.yaml
diff options
context:
space:
mode:
authorGustaf Rydholm <gustaf.rydholm@gmail.com>2022-06-09 22:36:04 +0200
committerGustaf Rydholm <gustaf.rydholm@gmail.com>2022-06-09 22:36:04 +0200
commit5c28ad1905f63838afbc7ee6c5ec31ff1310e3a1 (patch)
tree594796cc7d8bd01fedb167208dbf2593f2b898a1 /training/conf/network/conformer.yaml
parent4c7f720fe42278ec27e87bb1c3d5fcc6f1f822c2 (diff)
Move conformer config
Diffstat (limited to 'training/conf/network/conformer.yaml')
-rw-r--r--training/conf/network/conformer.yaml36
1 files changed, 36 insertions, 0 deletions
diff --git a/training/conf/network/conformer.yaml b/training/conf/network/conformer.yaml
new file mode 100644
index 0000000..1d72dd5
--- /dev/null
+++ b/training/conf/network/conformer.yaml
@@ -0,0 +1,36 @@
+_target_: text_recognizer.networks.conformer.Conformer
+depth: 8
+num_classes: 57
+dim: &dim 144
+dim_gru: 144
+block:
+ _target_: text_recognizer.networks.conformer.ConformerBlock
+ dim: *dim
+ attn:
+ _target_: text_recognizer.networks.conformer.Attention
+ dim: *dim
+ heads: 8
+ dim_head: 64
+ mult: 4
+ ff:
+ _target_: text_recognizer.networks.conformer.Feedforward
+ dim: *dim
+ expansion_factor: 4
+ dropout: 0.1
+ conv:
+ _target_: text_recognizer.networks.conformer.ConformerConv
+ dim: *dim
+ expansion_factor: 2
+ kernel_size: 31
+ dropout: 0.1
+subsampler:
+ _target_: text_recognizer.networks.conformer.Subsampler
+ pixel_pos_embedding:
+ _target_: text_recognizer.networks.transformer.AxialPositionalEmbedding
+ dim: 64
+ shape: [6, 127]
+ channels: 64
+ height: 6
+ dim: *dim
+ depth: 3
+ dropout: 0.1