nicholasKluge/Aira-2-124M

Partial fine-tune vs openai-community/gpt2. Partial change: 50% of the weights changed (attention, mlp, norms), in 6 of 12 layers.

Tensors changed: Norms 26/50, Attention 24/48, MLP 24/48, Embeddings 0/1

3 derivatives checked

• mradermacher/Aira-2-124M-i1-GGUF: Quantization
• mradermacher/Aira-2-124M-GGUF: Same weights, converted
• nicholasKluge/Aira-2-124M-DPO: Full fine-tune

On the Hub · Downloads on Model Pulse