llama : model-based max number of graph nodes

[ggerganov]

fix #8615

Propose to determine the max number of nodes based on the model info (arch, hparams, etc.)

• [x] I have read the contributing guidelines
• Self-reported review complexity:
  • [ ] Low
  • [ ] Medium
  • [ ] High

[ggerganov]

Ok, will merge this after the 405B model is release and the need for this change is confirmed. Likely the proposed n_layer > 400 check would have to be updated, because this number seems too big to me

[ceddybi]

@ggerganov what's left? up to now?

[ggerganov]

I haven't noticed any reports of 405B failing, so removed the increased max nodes limit for now and planning to merge just the new llama_model_max_nodes function