search

TheLastBen / fast-stable-diffusion

fast-stable-diffusion + DreamBooth
MIT License
7.48k stars 1.3k forks source link

Add stable diffusion 2 #599

Closed askiiart closed 1 year ago

askiiart commented 1 year ago

Stable Diffusion 2 was just released, it should probably be added sometime.

nawnie commented 1 year ago

OH MAN!!

x-legion commented 1 year ago

hell yeah!!

0xdevalias commented 1 year ago

What would your feature do ?

Support the new 768x768 model 2.0 from Stability-AI and all the other new models that just got released.

Links

See Also

0xdevalias commented 1 year ago

rudimentary support for stable diffusion 2.0

https://github.com/MrCheeze/stable-diffusion-webui/commit/069591b06bbbdb21624d489f3723b5f19468888d

Originally posted by @152334H in https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/5011#issuecomment-1325971596

GuruVirus commented 1 year ago

Looking forward to trying to run the new set of colabs 2.0 will need. The system reqs may be too large for free colab.

TheLastBen commented 1 year ago

On it!

nawnie commented 1 year ago

will be nice, you can train 768 on free its just time intensive and generate images, there will be alot more demand for the High ram setting i bet.

On Thu, Nov 24, 2022 at 1:39 AM Ben @.***> wrote:

On it!

— Reply to this email directly, view it on GitHub https://github.com/TheLastBen/fast-stable-diffusion/issues/599#issuecomment-1326063269, or unsubscribe https://github.com/notifications/unsubscribe-auth/AZPYLSABNXVP3K2JUYKZUCDWJ4LR5ANCNFSM6AAAAAASJTRPR4 . You are receiving this because you commented.Message ID: @.***>

-- shawn/nawnie

adamroszyk commented 1 year ago

Awesome, thanks again for all the great work you do!

0xdevalias commented 1 year ago

https://github.com/hafriedlander/diffusers/blob/stable_diffusion_2/scripts/convert_original_stable_diffusion_to_diffusers.py

Notes:

  • Only tested on the two txt2img models, not inpaint / depth2img / upscaling
  • You will need to change your text embedding to use the penultimate layer too
  • It spits out a bunch of warnings about vision_model, but that's fine
  • I have no idea if this is right or not. It generates images, no guarantee beyond that. (Hence no PR - if you're patient, I'm sure the Diffusers team will do a better job than I have)

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1326135768

Here's an example of accessing the penultimate text embedding layer https://github.com/hafriedlander/stable-diffusion-grpcserver/blob/b34bb27cf30940f6a6a41f4b77c5b77bea11fd76/sdgrpcserver/pipeline/text_embedding/basic_text_embedding.py#L33

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1326166368

doesn't seem to work for me on the 768-v model using the v2 config for v

TypeError: EulerDiscreteScheduler.init() got an unexpected keyword argument 'prediction_type'

Originally posted by @devilismyfriend in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1326220609

You need to use the absolute latest Diffusers and merge this PR (or use my branch which has it in it) https://github.com/huggingface/diffusers/pull/1386

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1326243809

(My branch is at https://github.com/hafriedlander/diffusers/tree/stable_diffusion_2)

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1326245339

0xdevalias commented 1 year ago

testing in progress on the horde https://github.com/Sygil-Dev/nataili/tree/v2 try it out Stable Diffusion 2.0 on our UI's

https://tinybots.net/artbot https://aqualxx.github.io/stable-ui/ https://dbzer0.itch.io/lucid-creations

https://sigmoid.social/@stablehorde/109398715339480426

SD 2.0

  • [x] Initial implementation ready for testing
  • [ ] img2img
  • [ ] inpainting
  • [ ] k_diffusers support

Originally posted by @AlRlC in https://github.com/Sygil-Dev/nataili/issues/67#issuecomment-1326385645

nawnie commented 1 year ago

ROCK ON!!! DO WE JUST NEED THE NEW CKPT?

On Thu, Nov 24, 2022, 6:53 AM Ben @.***> wrote:

Support added, requires more than 12GB of RAM (for now)

https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb

— Reply to this email directly, view it on GitHub https://github.com/TheLastBen/fast-stable-diffusion/issues/599#issuecomment-1326412740, or unsubscribe https://github.com/notifications/unsubscribe-auth/AZPYLSHRP5XDEDVT365FUKLWJ5QLNANCNFSM6AAAAAASJTRPR4 . You are receiving this because you commented.Message ID: @.***>

0xdevalias commented 1 year ago
TheLastBen commented 1 year ago

Should work now, make sure you check the box "redownload original model" when choosing V2

https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb

Requires more than 12GB of RAM for now, so free colab probably won't suffice.

djcedr commented 1 year ago

Awesome, thank you! I'm not sure I see how to train Dreambooth in the notebook - probably normal at this stage? Just checking :-)

TheLastBen commented 1 year ago

Awesome, thank you! I'm not sure I see how to train Dreambooth in the notebook - probably normal at this stage? Just checking :-)

For Dreambooth, we'll have to wait for the diffusers team to make it compatible with the V2 (probably very soon)

djcedr commented 1 year ago

Oh of course - thank you! I bumped into something though, I updated the repo, redownloaded the original model (v2), ran the requirements cell, then trying to start SD throws this:

LatentDiffusion: Running in v-prediction mode
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
DiffusionWrapper has 865.91 M params.
making attention of type 'vanilla-xformers' with 512 in_channels
building MemoryEfficientAttnBlock with 512 in_channels...
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla-xformers' with 512 in_channels
building MemoryEfficientAttnBlock with 512 in_channels...
^C

I'm on Colab Pro (not Pro+) and haven't pressed ^C :-) Thanks for your help

TheLastBen commented 1 year ago

In Colab : Runtime-> Change runtime type -> Runtime shape -> High-RAM

Keep the GPU Class to standard to save compute units

djcedr commented 1 year ago

Ah yes it worked, thanks!

Le jeu. 24 nov. 2022 à 19:06, Ben @.***> a écrit :

In Colab : Runtime-> change runtime type -> Runtime shape -> High-RAM

Keep the GPU Class to standard to save compute units

— Reply to this email directly, view it on GitHub https://github.com/TheLastBen/fast-stable-diffusion/issues/599#issuecomment-1326747515, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEKG55HDNYEQ5OPWRLYF4DWJ6VCTANCNFSM6AAAAAASJTRPR4 . You are receiving this because you commented.Message ID: @.***>

TheLastBen commented 1 year ago

@djcedr update the notebook again for a better V2

djcedr commented 1 year ago

Will do, thanks 🙏

Le jeu. 24 nov. 2022 à 20:15, Ben @.***> a écrit :

@djcedr https://github.com/djcedr update the notebook again for a better V2

— Reply to this email directly, view it on GitHub https://github.com/TheLastBen/fast-stable-diffusion/issues/599#issuecomment-1326786150, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEKG54J4KF62BFMRLWG733WJ65GRANCNFSM6AAAAAASJTRPR4 . You are receiving this because you were mentioned.Message ID: @.***>

0xdevalias commented 1 year ago

From @pcuenca on the HF discord:

We are busy preparing a new release of diffusers to fully support Stable Diffusion 2. We are still ironing things out, but the basics already work from the main branch in github. Here's how to do it:

  • Install diffusers from github alongside its dependencies:
pip install --upgrade git+https://github.com/huggingface/diffusers.git transformers accelerate scipy
  • Use the code in this script to run your predictions:
from diffusers import DiffusionPipeline, EulerDiscreteScheduler
import torch

repo_id = "stabilityai/stable-diffusion-2"
device = "cuda"

scheduler = EulerDiscreteScheduler.from_pretrained(repo_id, subfolder="scheduler", prediction_type="v_prediction")
pipe = DiffusionPipeline.from_pretrained(repo_id, torch_dtype=torch.float16, revision="fp16", scheduler=scheduler)
pipe = pipe.to(device)

prompt = "High quality photo of an astronaut riding a horse in space"
image = pipe(prompt, width=768, height=768, guidance_scale=9).images[0]
image.save("astronaut.png")

Originally posted by @vvvm23 in https://github.com/huggingface/diffusers/issues/1392#issuecomment-1326747275

0xdevalias commented 1 year ago

how sure are you that your conversion is correct? I'm trying to diagnose a difference I get between your 768 weights and my conversion script. There's a big difference, and in general I much prefer the results from my conversion. It seems specific to the unet - if I replace my unet with yours I get the same results.

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1327018829

OK, differential diagnostic done, it's the Tokenizer. How did you create the Tokenizer at https://huggingface.co/stabilityai/stable-diffusion-2/tree/main/tokenizer? I just built a Tokenizer using AutoTokenizer.from_pretrained("laion/CLIP-ViT-H-14-laion2B-s32B-b79K") - it seems to give much better results.

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1327031107

I've put "my" version of the Tokenizer at https://huggingface.co/halffried/sd2-laion-clipH14-tokenizer/tree/main. You can just replace the tokenizer in any pipeline to test it if you're interested.

Originally posted by @hafriedlander in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1327077503

GuruVirus commented 1 year ago

Should work now, make sure you check the box "redownload original model" when choosing V2

https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb

Requires more than 12GB of RAM for now, so free colab probably won't suffice.

Confirmed Free Colab crashes due to RAM capacity when loading the model (in another colab, not this one).

0xdevalias commented 1 year ago

@GuruVirus Have you tried this?

In Colab : Runtime-> Change runtime type -> Runtime shape -> High-RAM

Keep the GPU Class to standard to save compute units

Originally posted by @TheLastBen in https://github.com/TheLastBen/fast-stable-diffusion/issues/599#issuecomment-1326747515

0xdevalias commented 1 year ago

diffusers==0.9.0 with Stable Diffusion 2 is live!

https://github.com/huggingface/diffusers/releases/tag/v0.9.0

Originally posted by @anton-l in https://github.com/huggingface/diffusers/issues/1388#issuecomment-1327731012

0xdevalias commented 1 year ago

I've almost finished a proper implementation of Stable Diffusion 2.0 in Automatic1111, so that it runs locally and automatically updates everything and works on 4GB lowvram. It supports both 1.5 and 2.0 models and you can switch between models from the menu like normal.

So far the 512x512 base model, 512x512 inpainting model, and the 768x768 v-prediction model work properly. The upscaler model and depth models load correctly but don't work to generate images yet. It gives an error trying to load old Textual Inversion embeddings with the new models, but that can't be helped. And the PLMS sampling method isn't working. I'll push it soon.

Originally posted by @CarlKenner in https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/5011#issuecomment-1327367090

0xdevalias commented 1 year ago

when will Dreambooth support sd2

While it's not dreambooth, this repo seems to have support for finetuning SDv2:

Originally posted by @0xdevalias in https://github.com/JoePenna/Dreambooth-Stable-Diffusion/issues/112#issuecomment-1327993709

And looking at the huggingface/diffusers repo, there are a few issues that seem to imply people may be getting dreambooth things working with that (or at least trying to), eg.:

Originally posted by @0xdevalias in https://github.com/JoePenna/Dreambooth-Stable-Diffusion/issues/112#issuecomment-1327998619

0xdevalias commented 1 year ago

Issues potentially related to SDv2 dreambooth/finetuning/training/etc:

foshiz commented 1 year ago

Should work now, make sure you check the box "redownload original model" when choosing V2

https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast_stable_diffusion_AUTOMATIC1111.ipynb

Requires more than 12GB of RAM for now, so free colab probably won't suffice.

Dumb question: where does it spit out the URL for opening the webui after running everything?

askiiart commented 1 year ago

Dumb question: where does it spit out the URL for opening the webui after running everything?

At the last cell; it'll be either a loca.lt or gradio.app URL, I believe.

foshiz commented 1 year ago

Dumb question: where does it spit out the URL for opening the webui after running everything?

At the last cell; it'll be either a loca.lt or gradio.app URL, I believe.

Thanks. Odd, after running the last cell it's not returning either of those URL options in that google collab for me. It just returns the following after completing: LatentDiffusion: Running in v-prediction mode DiffusionWrapper has 865.91 M params. ^C

TheLastBen commented 1 year ago

Dumb question: where does it spit out the URL for opening the webui after running everything?

At the last cell; it'll be either a loca.lt or gradio.app URL, I believe.

Thanks. Odd, after running the last cell it's not returning either of those URL options in that google collab for me. It just returns the following after completing: LatentDiffusion: Running in v-prediction mode DiffusionWrapper has 865.91 M params. ^C

Out of RAM

foshiz commented 1 year ago

Dumb question: where does it spit out the URL for opening the webui after running everything?

At the last cell; it'll be either a loca.lt or gradio.app URL, I believe.

Thanks. Odd, after running the last cell it's not returning either of those URL options in that google collab for me. It just returns the following after completing: LatentDiffusion: Running in v-prediction mode DiffusionWrapper has 865.91 M params. ^C

Out of RAM

Dumb question: where does it spit out the URL for opening the webui after running everything?

At the last cell; it'll be either a loca.lt or gradio.app URL, I believe.

Thanks. Odd, after running the last cell it's not returning either of those URL options in that google collab for me. It just returns the following after completing: LatentDiffusion: Running in v-prediction mode DiffusionWrapper has 865.91 M params. ^C

Out of RAM

Oh got it. Thanks that worked.