Were there similar changes in 0.6.0? Reason I ask is, I upgraded to 0.6.0 from 0.5.7 and swarming quit working. I rolled back to 0.5.7 and it fixed the issue.
There may have been something. Did you report an error?
I created this thread to investigate any breaking changes in 0.6.0 that might have broken swarming code. We do have regression tests that run swarming, so if there were breaking changes, they were not covered in our tests. Looking at the release changelog, there are a couple of suspects:
- eliminate minPctActiveDutyCycle from spatial pooler
- Rename maxBoost to BoostStrength
No I didn’t report an error. I debugged it myself for an hour or so. And eventually I rolled my system back to 0.5.7, which resolved the issue. I wish I would have kept the output, but it was late at night and I was more concerned with getting it fixed.
I’m thinking it might be best for me to just upgrade straight to 0.7 at some point. But, if you think there’s value in me upgrading to 0.6.0 first, I’d be willing to give it a shot and report any errors. It’s your call though. My intention is to help find and stamp out any potential issues. But, I don’t want to create unnecessary work for anyone in the process.
@rhyolight Where should errors be reported?
Please report errors at https://github.com/numenta/nupic/issues/new
We won’t be able to do anything about it anyway, because 0.7.0-dev is full of breaking changes. I just want to make sure it is documented. If you had time to find what the error message was, that would help me with that. Then I could at least have a post to point to if/when others find the issue.
I’ll see if I can re-create the error. I’ll need to upgrade to 0.6 again in order to do so. Then again, since 0.7 is already released, maybe I’ll focus my efforts upgrading straight to 0.7 and watch for the same error.