Adding Matmul condition on srs size and cqlin setup assertions (Issue #136) #171
Conversation
Vimalakirti
left a comment
Vimalakirti
left a comment
There was a problem hiding this comment.
Thank you for your contributions! I left some comments.
added new empty line
There was a problem hiding this comment.
Thanks for contributing to our codebase! CQLinBB is actually used in multiple types of layers (e.g., Conv, Matmul, Conv2d). I think instead of directly embedding the logic into the Matmul layer I think it would be better to make a helper function with this logic and call it for each usage of CQLinBB. This would also make testing cleaner, as we would be able to test the function directly instead of copy-pasting the code in the test.
| let b = 64; | ||
| let matrix_size = a * b; | ||
|
|
||
| let input_vec_fr: Vec<Fr> = input_values.iter().map(|&x| f32_to_fr(x, 10)).collect(); |
There was a problem hiding this comment.
Can we simplify the test by generating a random Fr input and weights instead of loading floats from a file?
There was a problem hiding this comment.
Glad to contribute to the project. I will change this logic into a helper function to apply to the other layers as soon as I get some free time.
Regarding simplifying the test, I thought that loading from a file would be more realistic feeding directly the files but shouldn't be a concern if changed that way.
Will do.
3 participants
Added the constraints for issue #136 and tests