Open DarkAlchy opened 1 year ago
I mean, not necessarily. In order to do captioning with wd14 + danbooru + clipV2, yes. If you do one at a time, you can probably caption separately.
I mean, not necessarily. In order to do captioning with wd14 + danbooru + clipV2, yes. If you do one at a time, you can probably caption separately.
Nope, as I was just using clip and not enough ram with six gigs so I had someone show me and 11.7 to 17 gigs, but the clip is 11.7 gigs and I was told it is because you are loading up blip, and clip which is a lot of ram when I don't want the blip I only want the clip (V2). I could not use this at all because of it.
I did actually hit 17 gb usage with these settings earlier.
Crop to 768, Rename images Generate Captions Add Clip Results to Captions Use v2 Clip Model Append Flavor tags to CLIP (4) Upscale and Resize (2x res, SwinIR_4x)
Same here. When tried to use clip with flavours. For an instance CLIP Interrogator 0.5.1 uses VRAM check
detected < 12GB VRAM, using low VRAM mode
Didn't learn about that until after I went around to ask after trying it.
Needs to be said implicitly on the front page.