Debugging a run that will not learn on the How to Train Models track. Order: labels, loss falling on a tiny subset, gradients nonzero, then scale up. If a model cannot overfit 32 rows, the bug is in code, not capacity.
This lesson assumes you already worked through Seeds and repeated runs.
The idea in practice
Overfit a handful of batches on purpose. Then add data back.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'Debugging a run that will not learn',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
Changing architecture while the loss function reads the wrong column. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
List the four debug checks in order.
Self-check
- Say Debugging a run that will not learn in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.