CLRS-30 benchmark implementation and JAX compute kernel

[stjepanp]

Hi,

I’ve read the FloydNet paper and have been going through the repository to better understand the implementation.

I noticed that the part related to the CLRS-30 benchmark is currently missing. Would it be possible to include that section as well?

I’m particularly interested in:

• how the CLRS-30 benchmark integration was implemented, and
• how the “dedicated compute kernel” was realized in JAX within that setup.

Having access to this part of the code would make it easier to fully understand and reproduce the experimental pipeline.

[chivee]

Hi, @stjepanp

Thanks for your interest in our work.

We’re currently on a public holiday, and due to maintenance considerations, we’re still discussing how to gradually release CLRS-30 and the subsequent SALSA-CLRS components. One option we’re considering is forking DeepMind’s CLRS-30 as a separate standalone repository to keep the codebase clean and easier to maintain.

We will update this issue once the relevant parts are released.

Thanks for your patience.