>>105584853
i guess i need two load model flows so teacache doesnt apply to both sampling steps

this is not worth it, I could be genning during this time

unless you'd like to be a lad and patch OP Rentry's WAN workflow to use the two samplers but also use TeaCache. I also noticed you're doing DDIM but it seems like LCM is the one that is the new hotness (for self forcing especially)