From 6e9d4062f2e37240fefc5e04add8663c443c06bc Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Eren=20G=C3=B6lge?= <egolge@coqui.ai>
Date: Fri, 3 Sep 2021 13:20:26 +0000
Subject: [PATCH] Add `sort_by_audio_len` option

---
 TTS/tts/configs/shared_configs.py | 8 ++++++--
 TTS/tts/configs/vits_config.py    | 7 +++++--
 2 files changed, 11 insertions(+), 4 deletions(-)

diff --git a/TTS/tts/configs/shared_configs.py b/TTS/tts/configs/shared_configs.py
index 52e337f9..3dc70786 100644
--- a/TTS/tts/configs/shared_configs.py
+++ b/TTS/tts/configs/shared_configs.py
@@ -141,11 +141,14 @@ class BaseTTSConfig(BaseTrainingConfig):
         loss_masking (bool):
             enable / disable masking loss values against padded segments of samples in a batch.
 
+        sort_by_audio_len (bool):
+            If true, dataloder sorts the data by audio length else sorts by the input text length. Defaults to `True`.
+
         min_seq_len (int):
-            Minimum input sequence length to be used at training.
+            Minimum sequence length to be used at training.
 
         max_seq_len (int):
-            Maximum input sequence length to be used at training. Larger values result in more VRAM usage.
+            Maximum sequence length to be used at training. Larger values result in more VRAM usage.
 
         compute_f0 (int):
             (Not in use yet).
@@ -198,6 +201,7 @@ class BaseTTSConfig(BaseTrainingConfig):
     batch_group_size: int = 0
     loss_masking: bool = None
     # dataloading
+    sort_by_audio_len: bool = True
     min_seq_len: int = 1
     max_seq_len: int = float("inf")
     compute_f0: bool = False
diff --git a/TTS/tts/configs/vits_config.py b/TTS/tts/configs/vits_config.py
index 58fc66ee..39479231 100644
--- a/TTS/tts/configs/vits_config.py
+++ b/TTS/tts/configs/vits_config.py
@@ -67,11 +67,14 @@ class VitsConfig(BaseTTSConfig):
         compute_linear_spec (bool):
             If true, the linear spectrogram is computed and returned alongside the mel output. Do not change. Defaults to `True`.
 
+        sort_by_audio_len (bool):
+            If true, dataloder sorts the data by audio length else sorts by the input text length. Defaults to `True`.
+
         min_seq_len (int):
-            Minimum text length to be considered for training. Defaults to `13`.
+            Minimum sequnce length to be considered for training. Defaults to `0`.
 
         max_seq_len (int):
-            Maximum text length to be considered for training. Defaults to `500`.
+            Maximum sequnce length to be considered for training. Defaults to `500000`.
 
         r (int):
             Number of spectrogram frames to be generated at a time. Do not change. Defaults to `1`.