TianHao Zhang

What is the approximate duration of each audio in your dataset? If the input audios are very long, more GPU memory is required.

Are you using the same model structure and configuration as our repository provides? Whether to add or delete modules or change model size?

模型定义中的encoder.embed.out.0.weight与pre-trained model中的不同。 请检查模型的配置文件是否和pre-trained model一致以及 paddlespeech/s2t/modules/subsampling.py 中相关参数

请尝试一下develop版本是否会存在这个问题？

aishell/asr1默认是训练conformer，但从你错误信息中看是训练了deepspeech2，如果想训练deepseech2模型，可以参考aishell/asr0

Hi, is the missing data ready to be opened?

In my practical application, I can run spleeter on a single gpu as long as gpu version's tensorflow is installed correctly.

I encountered the same error "assert all_groups_norm > 0", does anyone have a solution?

Thank you! It works!