ML Environment Doctor v0.1.6¶

Diagnose 47+ ML environment signals in about 12 seconds.

mlenvdoctor finds the boring, expensive failures behind PyTorch, CUDA, Conda, Pipenv, Docker, TensorFlow, JAX, and LLM fine-tuning stacks before they waste your training run.

pip install mlenvdoctor
mlenvdoctor doctor --guided

Start diagnosing View checks

Terminal screenshot showing mlenvdoctor diagnosing CUDA, PyTorch, and Pipenv issues

PASS
CUDA driver visible

FAIL
PyTorch CUDA mismatch

WARN
Pipenv lock is stale

Quickstart¶

pip install mlenvdoctor          # 1. Install
mlenvdoctor doctor --guided      # 2. Get the next best fix
mlenvdoctor report               # 3. Generate JSON + HTML evidence

Why ML Engineers Use It¶

47+diagnostic signals across runtime, GPU, packages, and containers

12stypical fast triage for common local ML stacks

1 cmdactionable next step from `mlenvdoctor doctor --guided`

Built for the real failure loop

The goal is not just to print red text. mlenvdoctor groups low-level evidence into root-cause guidance: what failed, why it probably failed, what to do next, and how to verify it.

Common Workflows¶

Task	Command
Beginner-friendly recovery	`mlenvdoctor doctor --guided`
Automation-friendly triage	`mlenvdoctor doctor --json`
Full diagnostic table	`mlenvdoctor diagnose --full`
Machine-readable CI output	`mlenvdoctor doctor --ci`
JSON, CSV, or HTML evidence	`mlenvdoctor diagnose --json issues.json --html issues.html`
Safe fix plan	`mlenvdoctor fix --plan`
Dockerfile for training	`mlenvdoctor dockerize tinyllama`

Read the quickstart