diff options
Diffstat (limited to 'training/conf/network')
-rw-r--r-- | training/conf/network/decoder/transformer_decoder.yaml | 21 | ||||
-rw-r--r-- | training/conf/network/encoder/efficientnet.yaml | 6 |
2 files changed, 27 insertions, 0 deletions
diff --git a/training/conf/network/decoder/transformer_decoder.yaml b/training/conf/network/decoder/transformer_decoder.yaml new file mode 100644 index 0000000..60c5762 --- /dev/null +++ b/training/conf/network/decoder/transformer_decoder.yaml @@ -0,0 +1,21 @@ +_target_: text_recognizer.networks.transformer.Decoder +dim: 256 +depth: 2 +num_heads: 8 +attn_fn: text_recognizer.networks.transformer.attention.Attention +attn_kwargs: + num_heads: 8 + dim_head: 64 + dropout_rate: 0.2 +norm_fn: torch.nn.LayerNorm +ff_fn: text_recognizer.networks.transformer.mlp.FeedForward +ff_kwargs: + dim: 256 + dim_out: null + expansion_factor: 4 + glu: true + dropout_rate: 0.2 +rotary_emb: null +rotary_emb_dim: null +cross_attend: true +pre_norm: true diff --git a/training/conf/network/encoder/efficientnet.yaml b/training/conf/network/encoder/efficientnet.yaml new file mode 100644 index 0000000..1b9c6da --- /dev/null +++ b/training/conf/network/encoder/efficientnet.yaml @@ -0,0 +1,6 @@ +_target_: text_recognizer.networks.encoders.efficientnet.EfficientNet +arch: b0 +out_channels: 1280 +stochastic_dropout_rate: 0.2 +bn_momentum: 0.99 +bn_eps: 1.0e-3 |