Zackary Shen

If I guess right, you used `8` gpus to run the code. An annoying alternative is change `8` to `2`, making `3958` could divide `2` evenly . Hoping anyone could...

Hi @yenchenlin I have an observation that I got better result when I removed the `normalize_transforms` procedure. My scene is like an inward surrounding scene, the camera position differs a...

Try the latest commit?

**Updates**: Provide a python-interface implementation for distortion loss, no cuda kernels needed. ```python def einsum_distloss(w, m, interval): ''' Einsum realization of distortion loss. There are B rays each with N...

- Peak GPU memory (MB) | \# of pts `N` | 32 | 64 | 128 | 256 | 384 | 512 | 1024| |:------------:|:--:|:--:|:---:|:---:|:---:|:---:|:---:| |`original_distloss` |102|396|1560|6192|OOM|OOM|OOM| |`eff_distloss_native` |12|24|48|96|144|192|384| |`eff_distloss`...

@bchretien Yes, you are right. In my implementation I assume the sampling interval is identical for all rays. Because the uniform sampling is generally used for explicit voxel-based NERF methods,...

`torch.utils.cpp_extension` is for **`C++`** to create extension. However, `nms` module is written in **`C`** code here. So it doesn't work that you just replace `create_extension` to `BuildExtension` when import packages....

Maybe you can use version like these: `torch==0.4.1` `torchvision==0.2.2` `tensorboard==1.11.0 ` `tensorboardX==1.4`

> > Same question, I can download the videos without annotations. Where can I get the text(caption, annotation, transcription) data? Thanks a lot > > Hi, for ASR texts, please...

same issue.