netease-youdao/Confucius4-T3PO

Full fine-tune vs Qwen/Qwen2.5-14B. Fine-tune: 100% of the weights changed, across 48 of 48 layers.

Tensors changed: Embeddings 1/1, Norms 97/97, MLP 144/144, Attention 336/336, lm_head 1/1

2 derivatives checked

• netease-youdao/Confucius4-T3PO-GGUF: Same weights, converted
• mradermacher/Confucius4-T3PO-GGUF: Quantization

On the Hub · Downloads on Model Pulse