HIT-SCIR/huozi3.5

Full fine-tune vs HIT-SCIR/Chinese-Mixtral-8x7B. Fine-tune: 100% of the weights changed, across 32 of 32 layers.

Tensors changed: MLP 768/768, other 32/32, Norms 65/65, Attention 128/128

2 derivatives checked

• mradermacher/huozi3.5-i1-GGUF: Quantization
• mradermacher/huozi3.5-GGUF: Quantization

On the Hub · Downloads on Model Pulse