DavidAU/L3-DARKEST-PLANET-16.5B

Partial fine-tune (fewer layers) vs DavidAU/L3-Dark-Planet-8B. DavidAU/L3-Dark-Planet-8B matches part of DavidAU/L3-DARKEST-PLANET-16.5B (291 of its 642 tensors; the rest of DavidAU/L3-DARKEST-PLANET-16.5B has no counterpart). On the shared part: Partial change: 9% of the weights changed (attention, lm_head, mlp, norms), in 1 of 32 layers.

Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 3/65, MLP 3/96, Attention 4/128

3 derivatives checked

On the Hub · Downloads on Model Pulse