Davd-b01/lfm2.5-2.6B-thinkingcap-distiller

Partial fine-tune vs LiquidAI/LFM2.5-2.6B. Partial change: 90% of the weights changed (attention, embeddings, mlp, other), in 30 of 30 layers.

Tensors changed: Embeddings 1/2, other 22/66, MLP 90/90, Norms 0/76, Attention 32/32

On the Hub · Downloads on Model Pulse