<aside> ℹ️ At a glance Job: e.g. Loads and runs models, and computes the gradients evasion attacks use Learned in: M08 - AI Evasion - Foundations Install: e.g. pip install torch in a Python virtual env

</aside>

Go-To Commands

the ones you reach for first, best one on top

# command            # what it does

Options Worth Knowing

Option What it does When I use it
e.g. model.eval() e.g. Switches the model to inference mode e.g. Before every attack run

Use Cases

one heading per job you use it for

e.g. Check a model's prediction on a crafted input

Notes

Commands

# command            # what it does

Reading the Output

what a good result looks like, and what you do next

When It Fails

Symptom Likely cause Fix or backup tool
e.g. Results change every run e.g. Model left in training mode e.g. Call model.eval() first

Gotchas / Lessons Learned