What is the function of "Eros_beta6mix_Lora" ?
I used this Lora on beta6 and the video crashed
I too would like to know the purpose of this. Is it just the 30+ LoRA stack that was mentioned in the FastH3_VSA readme before it was labeled as unstable? Would be nice to know which LoRA's these have baked in so I'm not double stacking them. Sometimes it seems to know what I'm talking about without any LoRAs, other times they help, other times it feels overbaked. I really like how much memory these models save me with the LoRA's baked in.
People were extracting a difference to get something like it but that would pull a lot of support stuff and turbo deltas. The mix Lora is only Lora weights and tries to basically emulate what running the model but with control on strength, running it at 0.5 strength is probably the normal run. 1.0 seems to break it slightly. It's all the Lora components used in the model merge and their merged in the same ratios, but it's still a ranked Lora without the actual integration the merge used, so it'll never really be the same. It's basically just mix merge of all the nsfw Loras kind of. Using it ON beta6 is probably not a good idea.
People were extracting a difference to get something like it but that would pull a lot of support stuff and turbo deltas. The mix Lora is only Lora weights and tries to basically emulate what running the model but with control on strength, running it at 0.5 strength is probably the normal run. 1.0 seems to break it slightly. It's all the Lora components used in the model merge and their merged in the same ratios, but it's still a ranked Lora without the actual integration the merge used, so it'll never really be the same. It's basically just mix merge of all the nsfw Loras kind of. Using it ON beta6 is probably not a good idea.
@TenStrip did you merge any cinematic or realism lora into it? If so, do you remember which ones?