rizla/trrapi-16b

Partial fine-tune vs rizla/rizla-17. Partial change: 28% of the weights changed (attention, mlp, norms, other), in 11 of 39 layers.

Tensors changed: lm_head 0/1, MLP 66/234, other 11/39, Norms 22/79, Attention 44/156, Embeddings 0/1

1 derivatives checked

• mradermacher/trrapi-16b-GGUF: Quantization

On the Hub · Downloads on Model Pulse