Diagnostics CLI reference

areno env and areno check help diagnose setup problems before a user hits low-level Python, CUDA, or PyTorch errors.

areno env is a descriptive support report. It does not initialize the AReno engine or load model weights. Use it when collecting information for an issue.

areno env

For machine-readable issue reports:

areno env --json

The report includes:

  • AReno version

  • Python version and executable

  • OS, platform, and architecture

  • PyTorch version, CUDA build, CUDA runtime, and CUDA availability

  • CUDA driver information from nvidia-smi when available

  • visible GPU count, names, and compute capability

  • CUDA_HOME and inferred CUDA toolkit location

  • nvcc path and version

  • flash-attn import status and version

  • flash-linear-attention import status and version

  • areno_accel import status

  • selected environment variables such as MAX_JOBS, CUDA_VISIBLE_DEVICES, and TORCH_CUDA_ARCH_LIST

areno check

areno check validates whether the machine is ready to run AReno training and serving. It classifies each check as OK, WARN, or FAIL and prints concrete next steps for failures.

areno check

Example output:

AReno check: not ready

OK   Python >= 3.10
     found 3.11.8
OK   PyTorch CUDA build
     torch.version.cuda=12.4
OK   CUDA_HOME
     not set (not required for runtime; areno_accel imports)

CUDA_HOME and nvcc are only warnings when AReno needs to build its CUDA extension. If the installed areno_accel extension imports successfully, they are not required for runtime readiness.

Checks include:

  • Python version

  • supported platform

  • PyTorch import and version

  • PyTorch CUDA build

  • torch.cuda.is_available()

  • NVIDIA GPU visibility

  • CUDA_HOME and nvcc

  • optional runtime dependency imports

  • areno_accel import

  • writable cache/log locations

WARN items usually indicate degraded or incomplete setup. FAIL items mean AReno is not ready to run the CUDA training/inference engine.