Python Dependency Management for ML Quiz

Answers and explanations appear as you take the quiz.

1. What does a virtual environment primarily provide?
   • Isolation of project packages from other Python environments
   • A compatible host GPU driver
   • A guarantee of deterministic training
   • A lock on external datasets
2. What does a lockfile generally add beyond a dependency manifest?
   • A copy of the training dataset
   • A concrete resolved dependency set
   • A host kernel driver
   • A guarantee every wheel exists forever
3. Why can a generic framework install command choose an unsuitable GPU build?
   • Every machine has the same runtime libraries
   • Python packaging always ignores CUDA versions
   • Wheel variants depend on platform and accelerator compatibility
   • A lockfile sets host driver versions
4. What does a clean CI installation help detect?
   • All package vulnerabilities permanently
   • A model's causal validity
   • Future GPU availability
   • Undeclared dependencies available only on a developer machine
5. Which layer does a Python lockfile generally not control?
   • Transitive Python package versions
   • The host NVIDIA kernel driver
   • Some platform package markers
   • Resolved package hashes when supported
6. What should a team do when supporting both CPU and GPU installations?
   • Store the driver inside requirements.txt
   • Assume one wheel works identically on every platform
   • Test each supported environment and document package constraints
   • Avoid recording the framework version
7. Why use package hashes where appropriate?
   • They make dependency updates automatic
   • They force every package to use a GPU
   • They provide model accuracy estimates
   • They help verify downloaded artifacts match expected content
8. What does repeatable package resolution not guarantee by itself?
   • Bitwise-identical training across hardware and data pipelines
   • A recorded dependency graph
   • A clean project environment
   • A tested installation procedure