- Q4F : Q4_K feed-forawrd (Q5_1 for ffn_down due to shape constraints)
- Q8A : Q8_0 attention, Q8_0 output, Q8_0 embeds
- Q8SH : Q8_0 shared experts
Readable speeds on a 24GiB GPU + 64GB RAM
Readable speeds on a 24GiB GPU + 64GB RAM
有文件?提交种子。
提交种子zai-org/GLM-4.6V