emikulic

no profile record

comments

emikulic

3 года назадdiscuss

It's over for decels.

emikulic

3 года назадdiscuss

Three epochs means it sees each token three times. The dataset is ~1T like you said.

emikulic

3 года назадdiscuss

I want flying cars and moving sidewalks like I was promised by The Jetsons. :(

emikulic

3 года назадdiscuss

Try https://huggingface.co/ehartford/WizardLM-7B-Uncensored and related models. They're not even trained on smut, just the neutered responses were removed before the RLHF stage (IIUC)

emikulic

3 года назадdiscuss

I used to be bad at math but then I did a 360° on that.

emikulic

3 года назадdiscuss

Emad said SD cost $600,000 to train. I wonder if Midjourney also had to pay that to train from scratch.

emikulic

3 года назадdiscuss

What benefits does the Huggingface diffusers(?) implementation have over A1111?

emikulic

3 года назадdiscuss

I like to run A1111 in --api mode and write my own script to drive it over HTTP.

emikulic

3 года назадdiscuss

Wow, Reddit is going to lose a lot of pagerank.

emikulic

3 года назадdiscuss

You're thinking of archive.is

emikulic

3 года назадdiscuss

You mean the one that defeated two superpowers?

emikulic

3 года назадdiscuss

Last time I looked into this, the answer was "because huggingface transformers and torch.load are written to do it this way"

You could absolutely do something streaming, or mmap the weights instead of loading them into system RAM. Just the default interfaces don't.

emikulic

3 года назадdiscuss

This used to be the case but nowadays you can just pip install e.g. https://libraries.io/pypi/nvidia-cuda-runtime-cu11

emikulic

4 года назадdiscuss

I hope you will run your own locally instead of letting Midjourney block and/or mess with your prompts.

emikulic

4 года назадdiscuss

If edge devices includes my gamer PC then yes. For apps I'd recommend https://github.com/AUTOMATIC1111/stable-diffusion-webui