-
首先感谢老师分享这么好的项目,我在运行adaptive分支的时候,采用了项目提供的预训练模型进行微调训练,微调数据分别测试aishell3和magic data以及 D8,在执行tacotron_synthesize.py的时候,使用V100 GPU 单卡,生成梅尔频谱耗时 3秒,经过Griffin声码器耗时6秒(如果改用waveRNN耗时9秒),问题如下:
1.Tacotron2是否本身就…
-
HI, I get a error when` mfa train`
**`mfa train ./AISHELL3/ aishell3_dict.txt ./aishell_out/`**
INFO - Setting up corpus information...
INFO - Number of speakers in corpus: 218, average number of…
-
![image](https://user-images.githubusercontent.com/68454150/145553333-0e8dfc70-384f-4968-87bf-09f511243dd4.png)
-
Warning: you do not have any of the recognized datasets in D:\DATA\aidatatang_200zh\corpus\test.
The recognized datasets are:
LibriSpeech/dev-clean
LibriSpeech/dev-other
Li…
-
Traceback (most recent call last):
File "J:\ImitateSound\MockingBird-main\pre.py", line 76, in
create_embeddings(synthesizer_root=args.out_dir, n_processes=n_processes_embed, encoder_model_fp…
-
Is that possible train Chinese?
-
1. Describe the bug
"RuntimeError: Unable to find a valid cuDNN algorithm to run convolution" when I train aishell3 dataset using config "espnet/egs2/aishell3/tts1/conf/tuning/train_gst+xvector_con…
-
Arguments:
datasets_root: None
enc_models_dir: encoder\saved_models
syn_models_dir: synthesizer\saved_models
voc_models_dir: vocoder\saved_models
cpu: Fa…
-
我注意到你们在aishell3的样例中提到将stop token prediction去除,转用 attention 的最高点到达 encoder 侧的最后一个符号为终止条件,请问你们有验证过这个改动对效果的影响么?
-
看了关于aishell3的相关example来作为tts入门,学到了许多, 想了解下,如题中文中夹杂一些单词和字母, 该怎么形成如同simple.lexicon 以及包含拼音信息的 .lab文件呢 ,谢谢了