We are stuck because of multiple parallele trainings that need to be done, the task / job model let's us do that but we need one more trainer (VM for that) to be able to process more jobs in parallel also we need a clean way to recover in case the trainer VM breaks.
We should script the installation of donkeycar using Ansible for a paperspace basic ubuntu VM (P4000).
We are stuck because of multiple parallele trainings that need to be done, the task / job model let's us do that but we need one more trainer (VM for that) to be able to process more jobs in parallel also we need a clean way to recover in case the trainer VM breaks.
We should script the installation of donkeycar using Ansible for a paperspace basic ubuntu VM (P4000).