[TTS] deprecate AudioToCharWithPriorAndPitchDataset. (#5959)

Signed-off-by: Xuesong Yang <[email protected]>
NVIDIA · Feb 8, 2023 · 1703542 · 1703542
1 parent 4804f31
commit 1703542
Show file tree

Hide file tree

Showing 2 changed files with 1 addition and 19 deletions.
diff --git a/nemo/collections/tts/models/fastpitch.py b/nemo/collections/tts/models/fastpitch.py
@@ -95,19 +95,8 @@ def __init__(self, cfg: DictConfig, trainer: Trainer = None):
                 assert self.vocab is not None
                 input_fft_kwargs["n_embed"] = len(self.vocab.tokens)
                 input_fft_kwargs["padding_idx"] = self.vocab.pad
-            # TODO @xueyang: remove AudioToCharWithPriorAndPitchDataset because it has been deprecated already.
-            elif self.ds_class_name == "AudioToCharWithPriorAndPitchDataset":
-                logging.warning(
-                    "AudioToCharWithPriorAndPitchDataset class has been deprecated. No support for"
-                    " training or finetuning. Only inference is supported."
-                )
-                tokenizer_conf = self._get_default_text_tokenizer_conf()
-                self._setup_tokenizer(tokenizer_conf)
-                assert self.vocab is not None
-                input_fft_kwargs["n_embed"] = len(self.vocab.tokens)
-                input_fft_kwargs["padding_idx"] = self.vocab.pad
             else:
-                raise ValueError(f"Unknown dataset class: {self.ds_class_name}")
+                raise ValueError(f"Unknown dataset class: {self.ds_class_name}.")
 
         self._parser = None
         self._tb_logger = None
@@ -238,11 +227,6 @@ def parser(self):
 
             if ds_class_name == "TTSDataset":
                 self._parser = self.vocab.encode
-            elif ds_class_name == "AudioToCharWithPriorAndPitchDataset":
-                if self.vocab is None:
-                    tokenizer_conf = self._get_default_text_tokenizer_conf()
-                    self._setup_tokenizer(tokenizer_conf)
-                self._parser = self.vocab.encode
             else:
                 raise ValueError(f"Unknown dataset class: {ds_class_name}")
         else:

diff --git a/nemo/collections/tts/models/tacotron2.py b/nemo/collections/tts/models/tacotron2.py
@@ -133,8 +133,6 @@ def parser(self):
                 abbreviation_version="fastpitch",
                 make_table=False,
             )
-        elif ds_class_name == "AudioToCharWithPriorAndPitchDataset":
-            self.parser = self.vocab.encode
         else:
             raise ValueError("Wanted to setup parser, but model does not have necessary paramaters")