Skip to content

submit-aml

Submit jobs to Azure Machine Learning with minimal friction.

submit-aml wraps the azure-ai-ml SDK and provides two interfaces — a CLI and a Python API — for submitting training jobs, managing environments, mounting data, running sweeps, and more.

Key features

  • One-command job submission — submit training scripts to Azure ML with sensible defaults and minimal boilerplate.
  • CLI and Python API — use whichever interface fits your workflow.
  • Flexible environment management — use Docker images, build contexts with uv dependency resolution, conda environment files, or existing Azure ML environments.
  • Data mounting and downloading — attach Azure ML datasets and job outputs as inputs, and configure output datastores.
  • Hyperparameter sweeps — define grid sweeps inline with a concise parameter=[value1,value2] syntax.
  • Multi-node distributed training — scale to multiple nodes with MPI or PyTorch distributed, including GPU-aware configuration.
  • Built-in services — enable TensorBoard and VS Code remote debugging directly from CLI flags.
  • Layered configuration — set defaults via a TOML config file, environment variables, or CLI flags, with clear precedence rules.

Quick start

Install:

uv tool install submit-aml

submit-aml \
    --script train.py \
    --compute-target my-gpu-cluster \
    -- --learning-rate 1e-4

from submit_aml import submit_to_aml

submit_to_aml(
    compute_target="my-gpu-cluster",
    script_path="train.py",
    script_args=["--learning-rate", "1e-4"],
)

Next steps