nicholasKluge/Aira-2-355M

Partial fine-tune vs openai-community/gpt2-medium. Partial change: 25% of the weights changed (attention, mlp, norms), in 6 of 24 layers.

Tensors changed: Attention 24/96, Norms 26/98, MLP 24/96, Embeddings 0/1

3 derivatives checked

• mradermacher/Aira-2-355M-GGUF: Same weights, converted
• mradermacher/Aira-2-355M-i1-GGUF: Quantization
• Felladrin/gguf-sharded-Aira-2-355M: Quantization

On the Hub · Downloads on Model Pulse