Skip to main content

Configure Your First Run

Complete installation first. All paths below are relative to ~/magma-workspace.

Agent model​

Create agent.json in the workspace:

{
  "model": {
    "path": "/absolute/path/to/your/qwen-checkpoint",
    "format": "qwen",
    "quantization": "none"
  }
}

Replace the path with a compatible local checkpoint. This example uses the Qwen adapter without additional quantization, so the model must fit in available memory alongside simulation. Other model formats and quantization options are explained in Run your first LLM. Model weights are not included in the agent repository, and MAGMA checkpoints have not been released yet. See the beta status.

Generation configuration​

Create config.yaml in the same workspace:

magma_agent_address: "http://localhost:8888"
magma_planner_address: "http://localhost:8000"

backends: {}

coaching:
  enabled: false
  provider: llm

generate:
  seed: 42
  nb_branch: 1
  nb_env: 1
  nb_max_update: 1
  max_start_per_stage: 1
  randomized: false

The first run uses a provided preset, disables coaching, and passes --no-judge to run without an additional language-model backend. This checks installation, agent calls, and tool execution; it does not validate semantic answers or collect coached corrections. Scenario-defined physical success/error checks still apply. A model that runs correctly may still fail the task.

Always pass --config-path ./config.yaml in the example commands. Other settings come from the installed core defaults; do not edit files inside site-packages. When replacing a configuration block, provide all settings you want in that block.

Add a language-model backend later​

Coaching, semantic answer checks, and dynamic request generation may need a separate backend. This backend is not the MAGMA agent server. Once a compatible Ollama server and model are available, replace backends: {} with:

backends:
  local:
    type: ollama
    endpoint: "http://localhost:11434/v1/chat/completions"
    default_model: "YOUR_INSTALLED_MODEL"
    timeout: 60
    max_retry: 3
    headers:
      Content-type: "application/json"

Replace the model and URL with the actual service values. Both the generator and agent must be able to reach the backend. Remove --no-judge to enable answer verification; set coaching.enabled: true and remove --no-coaching when you want coaching. See coaching for the correction workflow and magma-gen run --help for available command-line options.