DATA-3461 - Refactoring code for classifications #4764
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
viambot
added
the
safe to test
viambot
added
safe to test
bhaney
requested changes
Jan 30, 2025
There was a problem hiding this comment.
Could you split this file into two, one with the functions that just have to do with classification, and the other the generic Tensor to protobuf conversion functions?
There was a problem hiding this comment.
of course! thank you for the first pass. I will update the PR description with all the manual testing
bhaney
reviewed
Jan 30, 2025
| } | ||
|
|
||
| // softmax takes the input slice and applies the softmax function. | ||
| func softmax(in []float64) []float64 { |
There was a problem hiding this comment.
I think you'll probably have to make this public, as I think it's used in the vision service
There was a problem hiding this comment.
I was able to get away with not exporting it. LMK if I'm missing something
viambot
added
safe to test
viambot
added
safe to test
viambot
added
safe to test
viambot
added
safe to test
bhaney
approved these changes
Feb 7, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In app, for the Cloud ML model service, we will need to have code that basically looks at flat tensors and then parses the relevant classifications from them. This is essentially duplicated in the vision service, since that it almost exactly what we do.
Please cross-reference this PR for a WIP of what we're trying to accomplish: https://github.com/viamrobotics/app/pull/7458
I am mainly looking for feedback on whether:
a) this refactoring makes sense
b) if this where all the code should live
c) if you want me to add any more tests to make sure I didn't break anything
Testing (02/07/2025)
I used the EfficientNet-ImageNet2012 model that's in production and configured a vision service locally using the changes in RDK (above) and compared against stable RDK (below)