Search Results
8/9/2025, 8:19:34 PM
>>106201367
>7900 xtx
Oh, that might be your problem more than anything else. I have a 4090 (Strix OC) and the performance hit is almost imperceptible with GGUF. Qwen Q8 also fits nicely into VRAM too with room to spare.
>>106203462
>doesn't EQ VAE help unscramble hands?
My current Jenny LoRA used EQ VAE during training and I haven't noticed any less shitty hands during inference, but I have noticed that there's less distortion in the likeness when the head/face is at extreme angles now. Leaning far over or even upside down no longer breaks the likeness.
>7900 xtx
Oh, that might be your problem more than anything else. I have a 4090 (Strix OC) and the performance hit is almost imperceptible with GGUF. Qwen Q8 also fits nicely into VRAM too with room to spare.
>>106203462
>doesn't EQ VAE help unscramble hands?
My current Jenny LoRA used EQ VAE during training and I haven't noticed any less shitty hands during inference, but I have noticed that there's less distortion in the likeness when the head/face is at extreme angles now. Leaning far over or even upside down no longer breaks the likeness.
Page 1