> "wanted to run glm-4.7-flash:q8_0"
> q8_0
a well made (as in, unsloth) smaller quant will help a good amount here, without a notable reduction in performance or increase in perplexity
a well made (as in, unsloth) smaller quant will help a good amount here, without a notable reduction in performance or increase in perplexity
No comments yet.