diffusers not working at all #1408
Description
LocalAI version:
2.0.0-ffmpeg
Environment, CPU architecture, OS, and Version:
Linux srv-gpt 5.15.131-1-pve #1 SMP PVE 5.15.131-2 (2023-11-14T11:32Z) x86_64 x86_64 x86_64 GNU/Linux, Proxmox LXC, AMD Ryzen 9 5900X, 128 GB RAM
Describe the bug
Diffusers is not working at all.
To Reproduce
I set up animagine-xl as stated in the docs. File "animagine-xl.yaml" in models folder with content:
parameters:
model: Linaqruf/animagine-xl
backend: diffusers
# Force CPU usage - set to true for GPU
f16: false
diffusers:
pipeline_type: StableDiffusionXLPipeline
cuda: false # Enable for GPU usage (CUDA)
scheduler_type: euler_a
Then ran command as stated in docs:
curl http://localhost:8080/v1/images/generations \
-H "Content-Type: application/json" \
-d '{
"prompt": "cat, outdoor, sun, tree|rain, night, people",
"model": "animagine-xl",
"step": 51,
"size": "1024x1024"
}'
It results always in this error:
{"error":{"code":500,"message":"could not load model (no success): Unexpected err=ValueError(\"Pipeline \u003cclass 'diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl.StableDiffusionXLPipeline'\u003e expected {'vae', 'scheduler', 'text_encoder_2', 'tokenizer', 'text_encoder', 'unet', 'tokenizer_2'}, but only {'vae', 'scheduler', 'tokenizer', 'text_encoder', 'unet'} were passed.\"), type(err)=\u003cclass 'ValueError'\u003e","type":""}}
In addition I tried another one (dreamlike-art/dreamlike-photoreal-2.0), which also fails with this error. So I assume, it does not work with any other model.
Also there is no files in /model folder, but disk size has increased, so it definitely downloaded something somewhere (also it loaded it into RAM as I saw in Proxmox). But where is it now stored? I want to delete it, because it uses diskspace.
Expected behavior
Diffusers generate images.
And models are stored in models folder too (or any other folder I can access to delete models I do not want anymore)