Nohobby/MS-Schisandra-22B-v0.3

Full fine-tune vs ArliAI/Mistral-Small-22B-ArliAI-RPMax-v1.1. Fine-tune: 100% of the weights changed, across 56 of 56 layers.

Tensors changed: Embeddings 1/1, Norms 59/113, MLP 168/168, Attention 224/224, lm_head 1/1

2 derivatives checked

• bartowski/MS-Schisandra-22B-v0.3-GGUF: Same weights, converted
• mradermacher/MS-Schisandra-22B-v0.3-i1-GGUF: Quantization

On the Hub · Downloads on Model Pulse