gemma3 with kvcache

[noringname]

Bug description

Great work! Just that I found the kvcache was not actually used in the infer part, did I miss something?

What operating system are you using?

None

Where do you run your code?

None

Environment

[rasbt]

Thanks for the note, and you are absolutely correct. I think I must have left in the wrong text generation function during testing. I just fixed it in the Olmo 3 from-scratch PR #914