DataRobot DRUM

1.17.15 verified Fri May 01 auth: no python

DRUM (DataRobot User Models) allows you to develop, test, and deploy custom models for the DataRobot platform. It supports Python 3.8–3.14 and is actively maintained. Releases follow a regular cadence aligned with DataRobot's platform updates.

pip install datarobot-drum

Common errors

error ModuleNotFoundError: No module named 'datarobot_drum' ↓

cause Package not installed or wrong import path.

fix

Run pip install datarobot-drum and use from datarobot_drum import ....

error ModuleNotFoundError: No module named 'datarobot.drum' ↓

cause Using old import path from version <1.0.

fix

Replace from datarobot.drum import ... with from datarobot_drum import ....

error ValueError: invalid literal for int() with base 10: 'some_string' ↓

cause Numeric class labels being treated as integers.

fix

Ensure target column contains string labels, not numeric.

error RuntimeError: DRUM does not support Python 3.15 ↓

cause Python version not supported.

fix

Use Python 3.8 to 3.14.

Warnings

breaking In v1.0, the package was renamed from `datarobot.drum` to `datarobot_drum`. All imports must be updated. ↓

fix Replace `from datarobot.drum import ...` with `from datarobot_drum import ...`.

deprecated In v1.6.5, the `--production` flag was deprecated. Use `--production` is not available for custom Flask extensions via `custom_flask.py`. ↓

fix Use `custom_flask.py` without `--production` flag to add Flask extensions.

gotcha Numeric class labels can be interpreted as numbers instead of strings. For classification, ensure labels are strings. ↓

fix Convert labels to strings in your custom model, e.g., `y = y.astype(str)`.

gotcha DRUM requires Python <3.15 and >=3.8. Using Python 3.15+ will cause installation failure. ↓

fix Ensure your Python version is between 3.8 and 3.14 inclusive.

Imports

RuntimeParameters

wrong

from datarobot.drum import RuntimeParameters

correct

from datarobot_drum import RuntimeParameters

Old package layout was inside datarobot namespace.

ModelBundle

wrong

from datarobot_drum.models import ModelBundle

correct

from datarobot_drum.model import ModelBundle

No 's' in model.

DrumServer

wrong

from datarobot.drum.server import DrumServer

correct

from datarobot_drum.server import DrumServer

Package renamed from 'datarobot.drum' to 'datarobot_drum' in v1.0.

Quickstart

Check DRUM import and basic pattern for custom model.

import datarobot_drum as drum
from datarobot_drum import RuntimeParameters

# Test DRUM installation
try:
    from datarobot_drum.model import ModelBundle
    print("DRUM is ready")
except ImportError as e:
    print(f"Import error: {e}")

# Minimal custom model example
# Create a file called custom.py with:
# def fit(X, y, row_weights=None, **kwargs):
#     pass
# def score(data, model, **kwargs):
#     return model.predict(data)

# Run with: drum fit --code-dir . --input training.csv --target target