search

aszc-dev / ComfyUI-CoreMLSuite

A set of custom nodes for ComfyUI that allow you to use Core ML models in your ComfyUI workflows.
GNU General Public License v3.0
92 stars 8 forks source link

Benchmarks #18

Open aszc-dev opened 7 months ago

aszc-dev commented 7 months ago

Some benchmark data showing comparison between safetensors MPS implementation and Core ML, presented in a table in readme could be helpful. It should contain machine info (Processor + RAM), conversion data (attention implementation), image size, compute unit, model used (SD1.5, SDXL, LCM, any one of them with LCM LoRA).

vbaros commented 6 months ago

Would be also nice to see how much more efficient is ANE comparing to GPU since running on a GPU eats the battery quite fast.

laurentVeliscek commented 2 months ago

Using SD 2.1, / coreML, it seems that ANE is 1.79 x faster than GPU with my macbook Air M3...

and coreML is about 25% faster in comfyUI using coreML Sampler SDXL vs SDXL Ksampler (efficient)

https://github.com/huggingface/swift-coreml-diffusers/issues/31