Open yyua8222 opened 1 year ago
I got the same error and it is duplicated issue as #263 and maybe #336. Please help me if there is anybody succeeds in training custom dataset.
Actually, I have fixed this problem, this should be the issue that the current CLIP version from huggingface has been updated, so just simply forced the CLIP initial model to the correct parameter size of the checkpoint, and it could run.
I solved it by update transformers
Hi, just trying for the script, but got an error when loading the pre-trained cliptextmodel, it seems that CLIPTextModel can not directly load as "openai/clip-vit-large-patch14", am I wrong and how can I fix this problem, thanks