Skip to content

Nengomaster #296

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 21 commits into from
Nov 27, 2020
Merged

Nengomaster #296

merged 21 commits into from
Nov 27, 2020

Conversation

hunse
Copy link
Collaborator

@hunse hunse commented Sep 10, 2020

Get working with Nengo master

@hunse hunse force-pushed the nengomaster branch 4 times, most recently from db5521b to 5343f7a Compare October 1, 2020 12:48
@hunse hunse mentioned this pull request Oct 14, 2020
1 task
Copy link
Member

@drasmuss drasmuss left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pushed some minor fixups, and then a few questions below.

@hunse
Copy link
Collaborator Author

hunse commented Oct 19, 2020

Fixups LGTM

Copy link
Collaborator Author

@hunse hunse left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good!

Nengo Bones and others added 6 commits November 18, 2020 19:01
@drasmuss drasmuss force-pushed the nengomaster branch 4 times, most recently from b3911b4 to 621dfb4 Compare November 19, 2020 20:00
@hunse hunse mentioned this pull request Nov 20, 2020
We are currently not set up to initialize voltages on the chip,
so better to keep this all zero.

Also:
- Use zero voltage for decode neurons (we could relax this in the
  future, but for now it gets things working the way they used to).
- Rename `step_math` to `step` so custom neuron types work.
- Fixed connection seeding.
These are not permitted in Nengo, so we now match that behaviour.
(We used to give a warning and convert to Dense.)
On NxSDK 0.9.8, multichip tests (using snips or not, i.e. either
value of `precompute`) have problems on Nahuku08.

Some Nahuku32 boards hang if started without the skip-power option,
so always pass that option when starting Nahuku32.

Also xfail multi-chip learning tests, because they sometimes fail
on the chip due to some sort of timeout, and the chip receiving
an unexpected shutdown signal.
Drop support for Nengo 3.0
hunse and others added 6 commits November 27, 2020 11:47
We previously had implemented this for createProcess, but we now
use createSnip everywhere.
Was unable to find a good weight exponent because the initial weights
were very small. This may have been a problem for a while, with a
lucky seed saving us from the error. Using larger initial weights
(by setting the transform) avoids the error.
@drasmuss drasmuss merged commit 6edc15c into master Nov 27, 2020
@drasmuss drasmuss deleted the nengomaster branch November 27, 2020 18:32
@hunse hunse mentioned this pull request Jan 18, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

3 participants