Di Chen

Hi @RaymondJiangkw , > I think it would be better to look at their paper. They use SIREN, which has been demonstrated to be as effective as positional encoding. This...

Hi @RaymondJiangkw Many thanks for your quick reply 😄 It's much clearer now. I just found [a repo applying SIREN to NeRF--](https://github.com/ventusff/improved-nerfmm), which may support the effectiveness of SIREN. In...

Yes, I find the same phenomenon when using distributed training. The only difference is that I removed the BatchNorm1d layer instead of writing a customized wrapper. I still haven't figured...

Many thanks for your feedback!

> Specifically, what is the best way to prepare mini-batches (including real time data augmentation) in order to keep the GPU busy? ``` python from numba import jit @jit(nogil=True) def...

> Hi @stevenhubhub. Did you try first to [`remove()`](https://loguru.readthedocs.io/en/stable/api/logger.html#loguru._logger.Logger.remove) the default handler? This should avoid duplicated output. Hi @Delgan, would you please be so kind as to add information about...

> @dichen-cd There is a note about this [in the Readme](https://github.com/Delgan/loguru#no-handler-no-formatter-no-filter-one-function-to-rule-them-all): > > > Note that you may also [`remove()`](https://loguru.readthedocs.io/en/stable/api/logger.html#loguru._logger.Logger.remove) a previously added handler by using the identifier returned while...

You have to provide additional information such as the dataset website, task description and containing content. `/data3` is a catalog for **public datasets**, means it's a place to share the...

Done. Yet be aware, `/data3` may not have enough free space for this dataset. Comment below if there's any problem.

Remember to update the size