Lue Fan

Here is the log [fsd_waymoD1_1x.log](https://github.com/tusen-ai/SST/files/9886382/fsd_waymoD1_1x.log). I delete the log you posted, which too long to post here. Could you upload it as a file?

https://github.com/tusen-ai/SST/blob/main/configs/fsd/fsd_waymoD1_1x.py#L236 The batchsize you use is too large. You should scale the number (pos samples limit) along with your batchsize (~ 128 * batchsize)

Could you please point out which part of config you have modified? It's not easy to carefully check the modification you have made according to your log. And why you...

1. It seems that you modify at least the number of GPUs according to your log, which changes the total batchsize and iterations. Since FSD use syncBN and a iteration-based...

We still use NMS in FSD. But FSD does not heavily rely on NMS. If I recall correctly, removing NMS lead to 2 AP performance drop in Vehicle class. This...

应该只需要改config里面的in_channel。

需要改很多和点云范围相关的config参数： 比如main config和data config里的point cloud range。main config里面的output shape，grid size，sparse shape。

@HobbitLong Thanks for your reply. I didn't make my point clearly. What confused me is that why your implementation is more efficient than infoNCE in CPC? I believe NCE tries...

@HobbitLong I seem to figure out why I'm confused. The `k` and `N` in Eq 1. and Eq 9. respectively in your paper mean the size of memory bank？ Does...

`height` means the vertical height of each laser beam. `incl` means the inclination of each laser beam. Note that the 64 laser beams of LiDAR are not emitted from a...