diff options
| author | Gustaf Rydholm <gustaf.rydholm@gmail.com> | 2021-08-08 19:59:55 +0200 |
|---|---|---|
| committer | Gustaf Rydholm <gustaf.rydholm@gmail.com> | 2021-08-08 19:59:55 +0200 |
| commit | 240f5e9f20032e82515fa66ce784619527d1041e (patch) | |
| tree | b002d28bbfc9abe9b6af090f7db60bea0aeed6e8 /training/conf/network | |
| parent | d12f70402371dda586d457af2a3df7fb5b3130ad (diff) | |
Add VQGAN and loss function
Diffstat (limited to 'training/conf/network')
| -rw-r--r-- | training/conf/network/decoder/pixelcnn_decoder.yaml (renamed from training/conf/network/encoder/pixelcnn_decoder.yaml) | 2 | ||||
| -rw-r--r-- | training/conf/network/decoder/vae_decoder.yaml | 2 | ||||
| -rw-r--r-- | training/conf/network/encoder/pixelcnn_encoder.yaml (renamed from training/conf/network/decoder/pixelcnn_encoder.yaml) | 2 | ||||
| -rw-r--r-- | training/conf/network/encoder/vae_encoder.yaml | 2 | ||||
| -rw-r--r-- | training/conf/network/vqvae.yaml | 2 | ||||
| -rw-r--r-- | training/conf/network/vqvae_pixelcnn.yaml | 2 |
6 files changed, 6 insertions, 6 deletions
diff --git a/training/conf/network/encoder/pixelcnn_decoder.yaml b/training/conf/network/decoder/pixelcnn_decoder.yaml index 3895164..cdddb7a 100644 --- a/training/conf/network/encoder/pixelcnn_decoder.yaml +++ b/training/conf/network/decoder/pixelcnn_decoder.yaml @@ -1,5 +1,5 @@ _target_: text_recognizer.networks.vqvae.pixelcnn.Decoder out_channels: 1 hidden_dim: 8 -channels_multipliers: [8, 8, 2, 1] +channels_multipliers: [8, 2, 1] dropout_rate: 0.25 diff --git a/training/conf/network/decoder/vae_decoder.yaml b/training/conf/network/decoder/vae_decoder.yaml index 0a36a54..a5e7286 100644 --- a/training/conf/network/decoder/vae_decoder.yaml +++ b/training/conf/network/decoder/vae_decoder.yaml @@ -1,5 +1,5 @@ _target_: text_recognizer.networks.vqvae.decoder.Decoder out_channels: 1 hidden_dim: 32 -channels_multipliers: [4, 4, 2, 1] +channels_multipliers: [8, 8, 4, 1] dropout_rate: 0.25 diff --git a/training/conf/network/decoder/pixelcnn_encoder.yaml b/training/conf/network/encoder/pixelcnn_encoder.yaml index 47a130d..f12957b 100644 --- a/training/conf/network/decoder/pixelcnn_encoder.yaml +++ b/training/conf/network/encoder/pixelcnn_encoder.yaml @@ -1,5 +1,5 @@ _target_: text_recognizer.networks.vqvae.pixelcnn.Encoder in_channels: 1 hidden_dim: 8 -channels_multipliers: [1, 2, 8, 8] +channels_multipliers: [1, 2, 8] dropout_rate: 0.25 diff --git a/training/conf/network/encoder/vae_encoder.yaml b/training/conf/network/encoder/vae_encoder.yaml index dacd389..58e905d 100644 --- a/training/conf/network/encoder/vae_encoder.yaml +++ b/training/conf/network/encoder/vae_encoder.yaml @@ -1,5 +1,5 @@ _target_: text_recognizer.networks.vqvae.encoder.Encoder in_channels: 1 hidden_dim: 32 -channels_multipliers: [1, 2, 4, 4] +channels_multipliers: [1, 2, 4, 8, 8] dropout_rate: 0.25 diff --git a/training/conf/network/vqvae.yaml b/training/conf/network/vqvae.yaml index d97e9b6..835d0b7 100644 --- a/training/conf/network/vqvae.yaml +++ b/training/conf/network/vqvae.yaml @@ -3,7 +3,7 @@ defaults: - decoder: vae_decoder _target_: text_recognizer.networks.vqvae.vqvae.VQVAE -hidden_dim: 128 +hidden_dim: 256 embedding_dim: 32 num_embeddings: 1024 decay: 0.99 diff --git a/training/conf/network/vqvae_pixelcnn.yaml b/training/conf/network/vqvae_pixelcnn.yaml index 10200bc..cd850af 100644 --- a/training/conf/network/vqvae_pixelcnn.yaml +++ b/training/conf/network/vqvae_pixelcnn.yaml @@ -5,5 +5,5 @@ defaults: _target_: text_recognizer.networks.vqvae.vqvae.VQVAE hidden_dim: 64 embedding_dim: 32 -num_embeddings: 512 +num_embeddings: 1024 decay: 0.99 |