Preeti
After converting the Stable Diffusion model to Core ML, you will obtain several files including SafetyChecker.mlmodelc, TextEncoder.mlmodelc, Unet.mlmodelc, VAEDecoder.mlmodelc, VAEEncoder.mlmodelc, merges.txt, and vocab.json. Due to the large size of these files, packaging them directly into the app is impractical, so a better solution is to have users download them within the app.
A .mlmodelc file is essentially a directory containing compiled model files and metadata for Core ML. Uploading this to static file storage means uploading a directory, not a single file. To download this virtual directory within the app, packaging it into a .tar file is recommended over a .zip file, as the extraction time for a .zip file is too long.
A .tar file is an archive file that combines multiple files and directories into a single file without compression. To package the files, use the command “tar -cvf VAEEncoder.tar VAEEncoder.mlmodelc” on a desktop system, and to extract them, use “tar -xvf VAEEncoder.tar”. Within the app, only the directory needs to be specified for extraction.
By using the .tar file format, the original 3.86GB file can be extracted in seconds, compared to a significantly longer extraction time for a .zip file that is only slightly smaller. This choice improves user experience by reducing extraction time to a few seconds of network transfer time.
Source link
Source link: https://medium.com/@codecola/how-to-efficiently-package-and-handle-stable-diffusion-model-files-78525d54f0ed?source=rss——stable_diffusion-5
GIPHY App Key not set. Please check settings