HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1

Full fine-tune vs mistral-community/Mixtral-8x22B-v0.1. Fine-tune: 100% of the weights changed, across 56 of 56 layers.

Tensors changed: Embeddings 1/1, MLP 1344/1344, other 56/56, Attention 224/224, Norms 72/113, lm_head 1/1

3 derivatives checked