search

Mikolaj / horde-ad

Higher Order Reverse Derivatives Efficiently - Automatic Differentiation library based on the paper "Provably correct, asymptotically efficient, higher-order reverse-mode automatic differentiation"
BSD 3-Clause "New" or "Revised" License
33 stars 6 forks source link

test: conv2d ad vs manual dInp and dKrn #74

Closed tranma closed 1 year ago

tranma commented 2 years ago

Hi @Mikolaj I've fixed the "manually differentiated" conv2d_dInp and dkrn and re-ran the tests against the ad version. Works 🎉 Apologies for the confusion.

Mikolaj commented 2 years ago

Thank you very much! Shall I review the PR, asking questions inline, or would you prefer me to wait for the meeting (in which case, I'd just merge, then improve the code to the best of my knowledge and ask for forgiveness during the meeting:)?

tranma commented 2 years ago

Yes we can just talk right here if you’d like. Feel free to review/comment inline

Mikolaj commented 1 year ago

@tranma: Thank you very much for the PR. Merging after CI passes.