huggingface / optimum-intel

🤗 Optimum Intel: Accelerate inference with Intel optimization tools

https://huggingface.co/docs/optimum/main/en/intel/index

Apache License 2.0

364 stars 101 forks source link

make unet and vae number of channels static #692

Closed eaidova closed 3 months ago

eaidova commented 3 months ago

What does this PR do?

uses static number of channels for unet and vae as this value is important for model architecture (number of channels for convolutions) and can not be changed in runtime and always expected to be the same like provided in config

HuggingFaceDocBuilderDev commented 3 months ago

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

eaidova commented 3 months ago

LGTM, why not adding it to optimum instead ?

@echarlaix, I'll open the same PR to optimum too if you agree, I just want unblock some work on openvino side where these changes can simplify demo workflow as soon as possible

echarlaix commented 3 months ago

LGTM, why not adding it to optimum instead ?

@echarlaix, I'll open the same PR to optimum too if you agree, I just want unblock some work on openvino side where these changes can simplify demo workflow as soon as possible

works for me thanks @eaidova