lucDoom

Results 3 issues of lucDoom

Hi,first thx for ur excellent work! when I use celebA lama-deep model to predict with 'refine=True', there always a problem : File "/DATA/jupyter/personal/project/cover_diff/lama/saicinpainting/evaluation/refinement.py", line 128, in _infer z1,z2 = forward_front(masked_image)...

I try to finetune 32bVL with 8*A800 ,and set --tune_mm_mlp True --tune_mm_llm True, I find that max_pixels cant be bigger than 224*224, any suggestion for how to make max_pixels bigger?...