Vionex-digital/Hakeem-7B

Full fine-tune vs tiiuae/Falcon-H1-7B-Instruct. Fine-tune: 95% of the weights changed, across 44 of 44 layers.

Tensors changed: Embeddings 0/1, MLP 132/132, Norms 106/133, other 208/308, Attention 176/176, lm_head 1/1

1 derivatives checked

• mradermacher/Hakeem-7B-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse