Unable to run launch script for bionemo

I followed the Quickstart Guide up to the launch code. This looks like this for me:

BIONEMO_IMAGE=nvcr.io/nvidia/clara/bionemo-framework:1.3 # Container with tag
BIONEMO_PATH=/workspace/bionemo # Location of code to be used /workspace/bionemo or /opt/nvidia/bionemo
PROJECT_PATH=$(pwd) # Path of env config and optional development code
DATA_PATH=${HOME}/data # Local path to save downloaded and processed data
RESULT_PATH=${HOME}/result/bionemo_experiments # Path for training results

WANDB_API_KEY= # WandB logging requires API key configuration

JUPYTER_PORT=8888 # Jupyter port for inference notebooks
PROJECT_MOUNT=/workspace/bionemo # Location of library in container /workspace/bionemo for dev work or /opt/nvidia/bionemo for non-dev use
REGISTRY=nvcr.io # Only required for registry login
REGISTRY_USER=“$oauthtoken” # Only required for registry login
NGC_CLI_API_KEY=$(grep -m 1 apikey ~/.ngc/config | head -n 1 | cut -d’ ’ -f3) # Requires NGC cli configuration

when I try to run it using bash launch.sh dev, nothing happens. The container does not start (I check with docker ps), there are no error messages.

  • the NGC_CLI_ORG is correct
  • permissions are given

Was could be the issue?

Dear @alenarzaeva

Thanks for your question.

I am assuming you have performed
docker pull nvcr.io/nvidia/clara/bionemo-framework:1.3


docker run -it --rm --gpus all nvcr.io/nvidia/clara/bionemo-framework:1.3 bash

commands, and you are inside the container while running subsequent tasks. The container status can be verified by running docker ps -a from a separate command line terminal.

Please follow the steps provided here: Quickstart Guide — NVIDIA BioNeMo Framework
and let us know if the issue persists.

Also, please make sure that your NGC_CLI_API_KEY is correctly set.