ShiZiqiang

> Hi Qinsheng, Have you solved the inf loss problem? Can you share the solution, since I have met the same problem?

> @ShiZiqiang Unfortunately no. I can not fix the issue. I gave up PyTorch and used jax. Maybe you are interested in [this repo](https://github.com/qsh-zh/gDDIM). Hi, Qinsheng, Thank you so much....

> @ShiZiqiang Unfortunately no. I can not fix the issue. I gave up PyTorch and used jax. Maybe you are interested in [this repo](https://github.com/qsh-zh/gDDIM). Hi, Qinsheng, Cannot wait to see...

您好，我记不太清了，印象中应该是有半个月的样子。我不太会用多卡训练，要是能够多卡训练，可能时间会缩短。如果您尝试出了如何用多卡训练，还请教教我。。。：）

Hi, e... i think the method can be used in multi-channel condition, but we haven't tried. Since Yi Luo has some work to apply DPRNN on multi-channel data. You can...

> > Hi, e... i think the method can be used in multi-channel condition, but we haven't tried. Since Yi Luo has some work to apply DPRNN on multi-channel data....

Hi, Thank you for your interest. You can check https://shiziqiang.github.io/tastas/ for samples. On Thu, Oct 14, 2021 at 7:00 PM shaoyating ***@***.***> wrote: > hi,thanks for sharing the code,could you...

Hi, it is basically 12GB and batch_size is 1. In our later experiments, we also used GPU with 24GB memory with batch_size 2.