[QST] How to configure per-tensor-scaling for NVFP4

[naruse-takamasa]

I am working with the NVFP4 data type.
From documentation, I understand that NVFP4 used both per-tensor-scaling and per-block-scaling, and that scaling values must be stored in memory for the hardware to use them.

However, I could not find any clear API or example showing how to actually set tensor-scaling values in memory before launching a GEMM kernel.

What is the correct way to store per-tensor-scaling values in memory ?

An example or reference would be very helpfull.

Thank you!

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[github-actions]

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.