Initial commit

configs/model/cfm/default.yaml

@@ -0,0 +1,3 @@
+name: CFM
+solver: euler
+sigma_min: 1e-4

configs/model/decoder/default.yaml

@@ -0,0 +1,7 @@
+channels: [256, 256]
+dropout: 0.05
+attention_head_dim: 64
+n_blocks: 1
+num_mid_blocks: 2
+num_heads: 2
+act_fn: snakebeta

configs/model/encoder/default.yaml

@@ -0,0 +1,18 @@
+encoder_type: RoPE Encoder
+encoder_params:
+  n_feats: ${model.n_feats}
+  n_channels: 192
+  filter_channels: 768
+  filter_channels_dp: 256
+  n_heads: 2
+  n_layers: 6
+  kernel_size: 3
+  p_dropout: 0.1
+  spk_emb_dim: 64
+  n_spks: 1
+  prenet: true
+
+duration_predictor_params:
+  filter_channels_dp: ${model.encoder.encoder_params.filter_channels_dp}
+  kernel_size: 3
+  p_dropout: ${model.encoder.encoder_params.p_dropout}

configs/model/matcha.yaml

@@ -0,0 +1,14 @@
+defaults:
+  - _self_
+  - encoder: default.yaml
+  - decoder: default.yaml
+  - cfm: default.yaml
+  - optimizer: adam.yaml
+
+_target_: matcha.models.matcha_tts.MatchaTTS
+n_vocab: 178
+n_spks: ${data.n_spks}
+spk_emb_dim: 64
+n_feats: 80
+data_statistics: ${data.data_statistics}
+out_size: null # Must be divisible by 4

configs/model/optimizer/adam.yaml

@@ -0,0 +1,4 @@
+_target_: torch.optim.Adam
+_partial_: true
+lr: 1e-4
+weight_decay: 0.0