[On Hold] Move nanoGPT tensors to L1 to speedup inference [Pending CI]

[Sudharsan-V]

• Store all intermediate tensors in the nanoGPT model to L1.
• Store all weight tensors in L1 (if they fit).

[vigneshkeerthivasanx]

This commit contains the following:

• Store all intermediate tensors in the nanoGPT model in L1.
• Store all supported weight tensors of nanoGPT model in L1.

Commit Link

[boris-drazic]

If this is ready for review can you make a PR?

[punithsekar]

This work is done and we are waiting for #4221 to get merged as this work is depended on it.