Adding option to do flow matching based duration prediction

configs/experiment/ljspeech_stoc_dur.yaml

@@ -0,0 +1,16 @@
+# @package _global_
+
+# to execute this experiment run:
+# python train.py experiment=multispeaker
+
+defaults:
+  - override /data: ljspeech.yaml
+  - override /model/duration_predictor: flow_matching.yaml
+
+# all parameters below will be merged with parameters from default configurations set above
+# this allows you to overwrite only specified parameters
+
+tags: ["ljspeech"]
+
+
+run_name: ljspeech

configs/model/duration_predictor/deterministic.yaml

@@ -0,0 +1,7 @@
+name: deterministic
+n_spks: ${model.n_spks}
+spk_emb_dim: ${model.spk_emb_dim}
+filter_channels: 256
+kernel_size: 3
+n_channels: ${model.encoder.encoder_params.n_channels}
+p_dropout: ${model.encoder.encoder_params.p_dropout}

configs/model/duration_predictor/flow_matching.yaml

@@ -0,0 +1,7 @@
+defaults:
+  - deterministic.yaml
+  - _self_
+
+sigma_min: 1e-4
+n_steps: 10
+name: flow_matching

configs/model/encoder/default.yaml

@@ -3,16 +3,8 @@ 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

@@ -1,6 +1,7 @@
 defaults:
   - _self_
   - encoder: default.yaml
+  - duration_predictor: deterministic.yaml
   - decoder: default.yaml
   - cfm: default.yaml
   - optimizer: adam.yaml