So I just captioned like a bunch of images with a job that ran for almost two days. I accidentally clicked combine interrogations after misunderstanding what it actually did. Now I have a bunch of images without caption files sharing their name due to this. Is there a way to load the database and make those files? OR am I screwed and need to re-run the job? It seems difficult to find more through documentation on this in webui docs or the git page from picobyte. Any advice is SUPER appreciated. #research-finetune #dataset-labeling #text-generation
#Working with db.json after using WD14
potent karma
delicate snow
So I just captioned like a bunch of images with a job that ran for almost two da...
almost two days? how many images are you captioning 
so what did the combine interrogations do exactly?
potent karma
400,000 images of dudes. I'm not 100%, but what I believe happened is that it would somehow point images with the exact same tags to the first image's caption text file. With as many photos as I ran, there were many images with the exact same caption, so this happened qute a few times.