before_start.md

Getting Started

we visualize our training details via wandb (https://wandb.ai/site).

visualization

1. you'll need to login

   $ wandb login

2. you'll need to copy & paste you API key in terminal

   $ https://wandb.ai/authorize

   or add the key to the "code/config/config.py" with

   C.wandb_key = ""

training

our code is trained with 4 x nvidia a100 gpus, alternatively, you can downsize the range view's input resolution to lower the training requirements, but please note that the final results cannot be guaranteed to match ours.

for training, please find the training scripts in "scripts" directory.

$ ./scripts/nuscenes_run.sh your_defined_labelled_num

$ ./scripts/kitti_run.sh your_defined_labelled_num

checkpoints

checkpoints and training logs are in this google drive link.

training details

some examples of training detail, please see this wandb link.

In details, after clicking the run (e.g., nuscenes_final_10_pct), you can checkout:

1. overall information (e.g., training command line, hardware information and training time).
2. training details (e.g., loss curves, validation results and visualization)
3. output logs (well, sometimes might crash ...)