The command successfully generates a folder of ml packages and a resources file.
I then proceed to use the Swift API for the processing resources(tried .all and CPU and NE) on MacOS. But the image generation failed with the following error:
TestingApp[3087:59450] [espresso] [Espresso::handle_ex_plan] exception=ANECF error: failed to load ANE model. Error=createProgramInstanceForModel:modelToken:qos:isPreCompiled:enablePowerSaving:statsMask:memoryPoolID:enableLateLatch:modelIdentityStr:error:: Program load failure (0x20004)
TestingApp[3087:59450] [coreml] Error plan build: -1.
TestingApp[3087:59450] [client] doUnloadModel:options:qos:error:: nil _ANEModel
TestingApp[3087:59450] [espresso] ANECF error:
TestingApp[3087:59158] [coreml] Error Domain=com.apple.CoreML Code=1 "For input feature 'images', the provided shape 1 × 512 × 512 × 3 is not compatible with the model's feature description." UserInfo={NSLocalizedDescription=For input feature 'images', the provided shape 1 × 512 × 512 × 3 is not compatible with the model's feature description., NSUnderlyingError=0x600002c20f60 {Error Domain=com.apple.CoreML Code=0 "MultiArray shape (1 x 512 x 512 x 3) does not match the shape (1 x 256 x 256 x 3) specified in the model description" UserInfo={NSLocalizedDescription=MultiArray shape (1 x 512 x 512 x 3) does not match the shape (1 x 256 x 256 x 3) specified in the model description}}}
I added debug logs in Python to print the size parameters (pipe unet), but found it to be the same as the Apple/StableDiffusion1-4
Hello,
I tried to use the following command to convert a stable diffusion model
rinna/japanese-stable-diffusion
to Apple Core ML package:The command successfully generates a folder of ml packages and a resources file.
I then proceed to use the Swift API for the processing resources(tried .all and CPU and NE) on MacOS. But the image generation failed with the following error:
I added debug logs in Python to print the size parameters (pipe unet), but found it to be the same as the Apple/StableDiffusion1-4
Replies are greatly appreciated!
Best,