[Docs] Update for InternLV3.5

[Bezdarnost]

📚 The doc issue

Hello, thank you for your incredible work!

Previously I always used your documentation to check needed amount of memory to tune the models.

Before it was easy for me to tune any your model with my setup(H200x8, LoRA). But now my training stuck without any error on the first iteration, when I try to tune InternVL3.5-241B-A28B(other models work well). How many memory do I need?

Suggest a potential alternative/fix

Add documentation about InternVL3.5