SCAIL 2. question on LoRA & compile
#5
by Privac - opened
Could you provide the already made INT8 of the SCAIL 2 model? If you are running an already baked INT8 how would you load a lora? By pre-load + on-the-fly Convrot & none or there's another method I’m missing? Is triton compile only possible on 3090 or am i doing something wrong? I remember being able to do so on .gguf files using a 3070 Ti, but now throws error. Thank you for making this node and models btw