YoungChanYY

> @rishikksh20 @anigi98932 In line 314 of fastspeech2.py, 3rd parameter ys is missed. you should insert ys in 3rd parameter. > > And I also found that hp.model.phoneme_acoustic_embed is missed...

many thanks!

出错的位置好像是在predict位置。当取消在训练过程中进行eval处理时，训练得以正常进行。大佬 Traceback (most recent call last): File "train_bart_text2abc.py", line 180, in main() File "train_bart_text2abc.py", line 163, in main model.train_model(train_df, eval_data=eval_df, split_on_space=True, matches=count_matches) File "textgen/seq2seq/bart_seq2seq_model.py", line 452, in train_model **kwargs, File...

多谢。 我看到另一处地方，应该有些问题： 在textgen/seq2seq/bart_seq2seq_utils.py的preprocess_data_bart(data)函数中，对target_ids 数据处理的问题和建议如下，大佬看看对不对。谢谢！ def preprocess_data_bart(data): input_text, target_text, tokenizer, args = data ...... target_ids = tokenizer.batch_encode_plus( [target_text], # max_length=args.max_seq_length, #原代码 max_length=args.max_length, #建议代码 padding="max_length", return_tensors="pt", truncation=True, )