migtissera/Trinity-2-Codestral-22B-v0.2

Full fine-tune vs mistralai/Codestral-22B-v0.1. Fine-tune: 100% of the weights changed, across 56 of 56 layers.

Tensors changed: Norms 91/113, MLP 168/168, Attention 224/224

2 derivatives checked

• bartowski/Trinity-2-Codestral-22B-v0.2-GGUF: Same weights, converted
• mradermacher/Trinity-2-Codestral-22B-v0.2-i1-GGUF: Quantization

On the Hub · Downloads on Model Pulse