tlphams/Wizard-Zephyr-Orpo-8x22B

Full fine-tune vs HuggingFaceH4/zephyr-orpo-141b-A35b-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 83/113, lm_head 1/1

2 derivatives checked

• mradermacher/Wizard-Zephyr-Orpo-8x22B-i1-GGUF: Quantization
• mradermacher/Wizard-Mixtral-8x22B-Instruct-v0.1-i1-GGUF: Quantization

1 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse