From 00ce973412384e92a44836f168de2a9a8827259c Mon Sep 17 00:00:00 2001 From: Mushroomcowisheggs <107208254+mushroomcowisheggs@users.noreply.github.com> Date: Sat, 18 Apr 2026 17:13:30 +0800 Subject: [PATCH] =?UTF-8?q?feat:=20=E6=B7=BB=E5=8A=A0=E6=95=B0=E6=8D=AE?= =?UTF-8?q?=E9=9B=86=E7=9A=84=E9=94=99=E8=AF=AF=E5=A4=84=E7=90=86=E6=8F=90?= =?UTF-8?q?=E7=A4=BA=20(#2758)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-authored-by: moomushroom <107208254+moomushroom@users.noreply.github.com> --- GPT_SoVITS/AR/data/dataset.py | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/GPT_SoVITS/AR/data/dataset.py b/GPT_SoVITS/AR/data/dataset.py index 402483d9..87b284d4 100644 --- a/GPT_SoVITS/AR/data/dataset.py +++ b/GPT_SoVITS/AR/data/dataset.py @@ -67,8 +67,10 @@ class Text2SemanticDataset(Dataset): ) ) # "%s/3-bert"%exp_dir#bert_dir self.path6 = semantic_path # "%s/6-name2semantic.tsv"%exp_dir#semantic_path - assert os.path.exists(self.path2) - assert os.path.exists(self.path6) + if not os.path.exists(self.path2): + raise FileNotFoundError(f"Phoneme data file not found: {self.path2}") + if not os.path.exists(self.path6): + raise FileNotFoundError(f"Semantic data file not found: {self.path6}") self.phoneme_data = {} with open(self.path2, "r", encoding="utf8") as f: lines = f.read().strip("\n").split("\n") @@ -131,7 +133,7 @@ class Text2SemanticDataset(Dataset): phoneme, word2ph, text = self.phoneme_data[item_name] except Exception: traceback.print_exc() - # print(f"{item_name} not in self.phoneme_data !") + print(f"Warning: File \"{item_name}\" not in self.phoneme_data! Skipped. ") num_not_in += 1 continue @@ -152,7 +154,7 @@ class Text2SemanticDataset(Dataset): phoneme_ids = cleaned_text_to_sequence(phoneme, version) except: traceback.print_exc() - # print(f"{item_name} not in self.phoneme_data !") + print(f"Warning: Failed to convert phonemes to sequence for file \"{item_name}\"! Skipped. ") num_not_in += 1 continue # if len(phoneme_ids) >400:###########2：改为恒定限制为semantic/2.5就行 @@ -228,7 +230,11 @@ class Text2SemanticDataset(Dataset): # bert_feature=torch.zeros_like(phoneme_ids,dtype=torch.float32) bert_feature = None else: - assert bert_feature.shape[-1] == len(phoneme_ids) + try: + assert bert_feature.shape[-1] == len(phoneme_ids) + except AssertionError: + print(f"AssertionError: The BERT feature dimension ({bert_feature.shape[-1]}) of the file '{item_name}' does not match the length of the phoneme sequence ({len(phoneme_ids)}).") + raise return { "idx": idx, "phoneme_ids": phoneme_ids,