QuantFunc

/

Minimax-H3-Quantfunc-4bit

Like 28

yes, i will share the workflow tomorrow

thanks

Is the model open source? Why does the current workflow require API keys? Will it be needed in the future? Will there still be updates to versions that do not require keys?

Get your own API key and use QuantFunc in ComfyUI completely free—or integrate the engine into your code with one of our SDK plans for performance that’s 50% faster than the ComfyUI plugin. See our pricing plans.

The WeChat QR code has expired

thx for remained,

我在SM75上面运行,比原来慢很多啊, 看看这是什么问题.

Which model were you using before? Also, you can try setting block_cache to 0.0.3.

btw you can turn the pinned_memory on

it will speed up the hold process when works with low vram like 8G~20G