256x256 image generation doesnt work

[axel578]

256x256 image generation doesnt work

[gpetters94]

Can you provide some more info? I'm not able to reproduce with runwayml/stable-diffusion-v1-5 - it seems to generate 256x256 images just fine.

[axel578]

Can you provide some more info? I'm not able to reproduce with runwayml/stable-diffusion-v1-5 - it seems to generate 256x256 images just fine.

Hey thanks for your reponse @gpetters94

The workflow that doesnt work for me is on a ryzen 7 7735 HS 680 M gpu,

workflow:
use https://civitai.com/models/9409/or-anything-v5ink as cpkt or safetensors

try to generate an image in 256x256 through the API in the api request.

If it works for you, could you tell me what you use as file config and command line arguments and API call arguments ?

[axel578]

Also custom model is not possible in API request

[axel578]

@gpetters94 any progress ? :)

[axel578]

@gpetters94 ??

[gpetters94]

I'm not able to reproduce what you're seeing, I've gotten that model to work just fine. Can you try these commands?

SHARK command: python3 apps/stable_diffusion/web/index.py --import_mlir --server_port 7860 --api --clear_all

API call: curl --header "Content-Type: application/json" --request POST --data '{"prompt":"Cat standing in a field", "negative_prompt": "deformed, ugly", "seed": "-1", "height": 256, "width": 256, "steps": 50, "cfg_scale": 7, "custom_vae": "https://civitai.com/models/9409/or-anything-v5ink"}' http://192.168.16.1:7860/sdapi/v1/txt2img

If those work, can you post the commands you were running? They might have been incorrectly formatted.