Seeds and repeated runs on the How to Train Models track. One seed can be lucky. For a close comparison, run three seeds and look at the spread. If the spread is larger than the gain, you did not improve the model.
This lesson assumes you already worked through Tracking runs.
The idea in practice
Set seeds and still expect small GPU noise. Report mean and spread.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'Seeds and repeated runs',
}
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
Publishing the best of twenty seeds as the method. 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
Say how many seeds you need before you believe a 0.2 point gain.
Self-check
- Say Seeds and repeated runs 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.