Error message

[silent light]

Can someone help me? I'm pretty new to all this.

[viscid plinth]

looks like you use to much of your vram

[viscid plinth]

@silent light you can write here and dont need to dm me.
the error says that your vram is full. do you have change your invokeai.yaml ?

[silent light]

I have no idea how to change or alter anything, I was hoping someone could help to explain it to me.

[viscid plinth]

did generate any image with the settings or is this the first generation with the model?

[silent light]

0No, first try I get message

[viscid plinth]

can you try the any another flux model out of the starter kit but not the full models only the Quantization versions

[silent light]

I ran tests last night. It's only the off site FLux that will not work. Dev Quantization works fast and perfect. Nothing from Civitai will. SDXL and SD-1 models both work, but take a long time to make initial picture. If you puut in for a batch of say three? The first one takes foever, but eventually goes, then the next two are fast as can be.

[viscid plinth]

with only 16gb of vram you can only run Quantization models. the slow starts is because the models needs to be loaded into vram which can be slow from a spinning disk. from a ssd should be faster.

[silent light]

I read that I can do other thingst ohelp? Batch sizes, images sizes? model size, or is that the quantization?

[viscid plinth]

the models size with quantization is lower.
image size sure can help but with 512x512 i would dont go lower then that.

[silent light]

ok, ty so much for all the info!

[silent light]

oh! One last question please? Can you recommend a place to get the Quantization models for flux? There doesn't seem to be many I can find.

[viscid plinth]

i would say huggingface or civitai and look at the model size before the download