nicholasKluge/Aira-2-portuguese-560M

Partial fine-tune vs bigscience/bloom-560m. Partial change: 25% of the weights changed (attention, mlp, norms), in 6 of 24 layers.

Tensors changed: MLP 24/96, Norms 26/98, Attention 24/96, Embeddings 0/2

2 derivatives checked

• mradermacher/Aira-2-portuguese-560M-i1-GGUF: Quantization
• mradermacher/Aira-2-portuguese-560M-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse