Zyphra/Zamba2-2.7B-instruct

Full fine-tune vs Zyphra/Zamba2-2.7B. Fine-tune: 100% of the weights changed, across 54 of 54 layers.

Tensors changed: Embeddings 1/1, Norms 30/113, other 264/387, MLP 22/22, Attention 8/8

1 derivatives checked

• EchoLabs33/zamba2-2.7b-instruct-hxq: Same weights, converted

On the Hub · Downloads on Model Pulse