appvoid/arco-plus

Partial fine-tune (fewer layers) vs appvoid/arco. appvoid/arco matches part of appvoid/arco-plus (147 of its 237 tensors; the rest of appvoid/arco-plus has no counterpart). On the shared part: Partial change: 20% of the weights changed (attention, lm_head, mlp, norms), in 2 of 16 layers.

Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 5/33, MLP 6/48, Attention 8/64

1 derivatives checked

On the Hub · Downloads on Model Pulse