Support for LiquidAI/LFM2-1.2B

[JoseGhDark-commits]

Hi! It would be amazing to have support for LiquidAI, their models are very good in relation to their hardware requirements and their responses as an AI model.

[rimmer]

I wonder if we can just convert weights from LFM2

[oestape]

LFM2-8B-A1B is my favorite. It is a Mixture-of-Experts (MoE) with 8.3B total parameters and 1.5B active parameters per token. The answers are of good quality, similar to qwen3:8b, but much faster. It runs fast even in CPU, I wonder how it would go with webllm in a smartphone.