How can I make NVIDIA nvrtc available on my VM?

I have a repository which uses CUDA.

In the past, I only used the CUDA Runtime API, and was able to make it available to me using some copy-and-paste’ing from other projects’ .travis.yml (see below). Now, I also need to use CUDA’s nvrtc library. I know it’s supposed to be part of the CUDA distribution - but somehow it’s missing during the build on the VMs. What do I do?

More generally - is there a way to get a shell session on one of these VMs so as to “poke around” and check what’s available and what’s not using apt, find, and similar tools? If I could do that maybe I could figure things out on my own.

    - name: CUDA 9.2
      - CUDA=9.2.148-1
      - CUDA_SHORT=9.2
      - CUDA_APT=9-2
      - UBUNTU_VERSION=ubuntu1604
      dist: xenial
    - name: CUDA 10.1
      - CUDA=10.1.105-1
      - CUDA_APT=10-1
      - CUDA_SHORT=10.1
      - UBUNTU_VERSION=ubuntu1804
      dist: bionic
    - name: CUDA 10.2
      - CUDA=10.2.89-1
      - CUDA_APT=10-2
      - CUDA_SHORT=10.2
      - UBUNTU_VERSION=ubuntu1804
      dist: bionic

  - CUDA_REPO_PKG=cuda-repo-${UBUNTU_VERSION}_${CUDA}_amd64.deb
  - wget${UBUNTU_VERSION}/x86_64/${CUDA_REPO_PKG}
  - sudo dpkg -i ${CUDA_REPO_PKG}
  - wget${UBUNTU_VERSION}/x86_64/
  - sudo apt-key add
  - sudo apt update -qq
  - sudo apt install -y  cuda-core-${CUDA_APT} cuda-cudart-dev-${CUDA_APT} cuda-nvtx-${CUDA_APT}
  - sudo apt clean
  - CUDA_HOME=/usr/local/cuda-${CUDA_SHORT}
  - PATH=${CUDA_HOME}/bin:${PATH}