LESSON 0.6 · Lab · 120 min
PyTorch tensors and reproducible experiments
PyTorch tensors and reproducible experiments. Learn dtype, device, strides, autograd, and random seeds and complete: Build a reproducible tensor-lab repository. Part of the “Before
Learning objectives
- Explain what problem “PyTorch tensors and reproducible experiments” solves without hiding behind terminology.
- Trace the variables and causal links across dtype, device, strides.
- Complete “Build a reproducible tensor-lab repository” and judge the result with evidence rather than intuition.
Core concepts
dtype
dtype is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment before moving to an optimized implementation.
device
device is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
strides
strides is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
autograd
autograd is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
and random seeds
and random seeds is part of the lesson’s causal model. State its inputs, outputs, invariants, and failure mode; then verify it with a hand-check or a minimal experiment.
Build and verify
Build a reproducible tensor-lab repository
- Predict: write the expected output, trend, or failure before running code.
- Build: implement only the minimum components needed to answer the question.
- Verify: compare with a baseline or trusted implementation; save seeds, parameters, and raw outputs.
- Transfer: change one shape, dataset, scale, or workload condition and explain whether the conclusion still holds.