How can I train from scratch？

[Rogerlv51]

hello, thank for your outstanding work! I wonder how can I train from scratch by using SemanticKITTI datasets? Hope your reply, thank u. @Sangminhong

[Sangminhong]

hello, thank for your outstanding work! I wonder how can I train from scratch by using SemanticKITTI datasets? Hope your reply, thank u. @Sangminhong
You can try this command.
CUDA_VISIBLE_DEVICES=0 python main.py --experiment_id {experiment id} --dataset_name {SemanticKITTI} --class_name {car}

If it still doesn't works, please tell me the detailed issue you have. Thank you

[Rogerlv51]

Thanks for your reply! However, when I building the virtual environment, i have met this problems:
Pip subprocess error:
ERROR: Could not find a version that satisfies the requirement chamfer-3d==0.0.0 (from versions: none)
ERROR: No matching distribution found for chamfer-3d==0.0.0

failed

CondaEnvException: Pip failed

I just used your command: conda env create -f environment.yml --name ACL_SPC @Sangminhong

[Sangminhong]

Thanks for your reply! However, when I building the virtual environment, i have met this problems: Pip subprocess error: ERROR: Could not find a version that satisfies the requirement chamfer-3d==0.0.0 (from versions: none) ERROR: No matching distribution found for chamfer-3d==0.0.0

failed

CondaEnvException: Pip failed

I just used your command: conda env create -f environment.yml --name ACL_SPC @Sangminhong

Sorry that the environment.yml file did not work for you. Since every computer has different settings, it is difficult to make environment.yml that works for everyone. In this case you can download the packages by yourself which is not very difficult. If you still have problems, please feel free to ask.