3 ms·

Not M1 comparible but I'm working on testing various GPU vs M1 comparisons, with a few accessible cloud providers. My impression is times should be the same, bu

Not M1 comparible but I'm working on testing various GPU vs M1 comparisons, with a few accessible cloud providers. My impression is times should be the same, but it's nice to hear other real-world stats for M1 with SD. Makes me really want to rent the Hetzner M1 now. Which repo or build are you using BTW, is it the one related to this readme? https://github.com/magnusviri/stable-diffusion/blob/main/README-Mac-MPS.md https://github.com/magnusviri/stable-diffusion/blob/main/REA...

mrtksn 4y ago
>Which repo or build are you using BTW, is it the one related to this readme? https://github.com/magnusviri/stable-diffusion/blob/main/REA https://github.com/magnusviri/stable-diffusion/blob/main/REA... Yes, this one. However it was like a month ago I think, so speeds might have improved. I'm getting ~2.2s/step with another implementation: https://news.ycombinator.com/item?id=33006447 https://news.ycombinator.com/item?id=33006447
- thisisjasononhn 4y ago
  Wow, that sounds like a good improvement. I am also wondering, do you follow the general advice of 1 iteration and 1 sample, for example: --n_samples 1 --n_iter 1 (when referencing commands using txt2img.py) I figure you could wait a bit for things to process going further, but curious just if you're getting results like that with higher sample/iter settings.
  - mrtksn 4y ago
    I usually go with the default parameters.
stared 4y ago
I would love to see it, but this file is not accessible.
- thisisjasononhn 4y ago
  Sorry about that, web link rot sure is real eh. This is an example of the original file: https://github.com/magnusviri/stable-diffusion/blob/79ac0f34201b69c4e8d44ef476a6d96a7844c900/README-Mac-MPS.md https://github.com/magnusviri/stable-diffusion/blob/79ac0f34... Which seems to have been renamed, and cleaned up a bit here: https://github.com/magnusviri/stable-diffusion/blob/main/docs/installation/INSTALL_MAC.md https://github.com/magnusviri/stable-diffusion/blob/main/doc... However, per the note on the magnusviri repo, the following repo should be used for a stable set of this SD Toolkit: https://github.com/invoke-ai/InvokeAI https://github.com/invoke-ai/InvokeAI with instructions here https://github.com/invoke-ai/InvokeAI/blob/main/docs/installation/INSTALL_MAC.md https://github.com/invoke-ai/InvokeAI/blob/main/docs/install...

mrtksn 4y ago

>Which repo or build are you using BTW, is it the one related to this readme? https://github.com/magnusviri/stable-diffusion/blob/main/REA https://github.com/magnusviri/stable-diffusion/blob/main/REA... Yes, this one. However it was like a month ago I think, so speeds might have improved. I'm getting ~2.2s/step with another implementation: https://news.ycombinator.com/item?id=33006447 https://news.ycombinator.com/item?id=33006447

thisisjasononhn 4y ago
Wow, that sounds like a good improvement. I am also wondering, do you follow the general advice of 1 iteration and 1 sample, for example: --n_samples 1 --n_iter 1 (when referencing commands using txt2img.py) I figure you could wait a bit for things to process going further, but curious just if you're getting results like that with higher sample/iter settings.
- mrtksn 4y ago
  I usually go with the default parameters.

stared 4y ago

I would love to see it, but this file is not accessible.

thisisjasononhn 4y ago
Sorry about that, web link rot sure is real eh. This is an example of the original file: https://github.com/magnusviri/stable-diffusion/blob/79ac0f34201b69c4e8d44ef476a6d96a7844c900/README-Mac-MPS.md https://github.com/magnusviri/stable-diffusion/blob/79ac0f34... Which seems to have been renamed, and cleaned up a bit here: https://github.com/magnusviri/stable-diffusion/blob/main/docs/installation/INSTALL_MAC.md https://github.com/magnusviri/stable-diffusion/blob/main/doc... However, per the note on the magnusviri repo, the following repo should be used for a stable set of this SD Toolkit: https://github.com/invoke-ai/InvokeAI https://github.com/invoke-ai/InvokeAI with instructions here https://github.com/invoke-ai/InvokeAI/blob/main/docs/installation/INSTALL_MAC.md https://github.com/invoke-ai/InvokeAI/blob/main/docs/install...