Calc-X¶

GPU	Model	Controller Mode	Trainer Mode	Code
1× A100 80GB	`Qwen/Qwen2.5-1.5B-Instruct`	K8s or local	Sync and async	Source

Calc-X is a proof-of-concept (POC) example that trains a mathematical reasoning agent on the Calc-X dataset with verl and Agent Lightning >=v1.0. It is intentionally lightweight and requires only one GPU. The agent uses AutoGen + MCP calculator tools to solve math problems.

The example supports two controller modes:

K8s mode: Minikube provides a minimal Kubernetes environment, and agent rollouts run as Kubernetes Jobs.
Local mode: Agent rollouts run directly as local processes without Kubernetes.

Both synchronous and asynchronous trainer modes are supported.

Data Preparation¶

Download the Calc-X dataset from Google Drive, then extract it into examples/calc_x/data/:

cd examples/calc_x
unzip data/calc-x-data.zip -d data/

The expected dataset files are:

data/train.parquet
data/test.parquet
data/test_mini.parquet
data/sample.jsonl

Local Mode¶

Make sure you have activated the project environment and installed the following package in Python:

source .venv/bin/activate
uv pip install \
    openai \
    httpx \
    sympy \
    "autogen-agentchat" \
    "autogen-ext[openai]" \
    "mcp>=1.11.0,<2" \
    mcp-server-calculator

Then start training:

source .venv/bin/activate
cd examples/calc_x
bash run_local.sh

run_local.sh starts agl-server and agl-controller, and writes their logs under /tmp/. The script starts the agent in multi-process mode. When run_local.sh exits, it automatically cleans up the server, controller, and agent it started.

K8s Mode¶

This example uses Minikube to demonstrate the minimal Kubernetes workflow. For production deployments, replace Minikube with a production-grade Kubernetes cluster.

Make sure you have installed docker and minikube, then start training by:

source .venv/bin/activate
cd examples/calc_x
bash run_minikube.sh

run_minikube.sh starts agl-server and agl-controller, and writes their logs under /tmp/. The script also starts a new local Minikube single-node K8s cluster, and the agent runs in this cluster as Kubernetes Jobs. When run_minikube.sh exits, it automatically cleans up the server, controller, and Minikube it started. Minikube needs at least 64 GB of memory; otherwise, it may be killed due to insufficient memory.