olaverse/MIST-1-140B

Partial fine-tune (fewer layers) vs olaverse/MIST-1-70B. olaverse/MIST-1-70B matches part of olaverse/MIST-1-140B (723 of its 1425 tensors; the rest of olaverse/MIST-1-140B has no counterpart). On the shared part: Partial change: 1% of the weights changed (attention, mlp, norms), in 1 of 80 layers.

Tensors changed: lm_head 0/1, Embeddings 0/1, Norms 2/161, MLP 3/240, Attention 4/320

2 derivatives checked

On the Hub · Downloads on Model Pulse