| license: creativeml-openrail-m | |
| tags: | |
| - core-ml | |
| - stable-diffusion | |
| - text-to-image | |
| base_model: nota-ai/bk-sdm-tiny-2m | |
| library_name: ml-stable-diffusion | |
| # CoreML BK-SDM-Tiny (fp16, ORIGINAL attention) | |
| Apple **Core ML** conversion of [nota-ai/bk-sdm-tiny-2m](https://huggingface.co/nota-ai/bk-sdm-tiny-2m) — | |
| a block-removed, knowledge-distilled Stable Diffusion (SD 1.4-class, ~0.33B U-Net). | |
| - **Format:** `original/compiled` (GPU / `.cpuAndGPU`), fp16, for Apple's | |
| [ml-stable-diffusion](https://github.com/apple/ml-stable-diffusion) Swift pipeline. | |
| - **Resolution:** 512x512 · **Steps:** ~10 (distilled) · standard scheduler + CFG. | |
| - **Size:** ~0.95 GB (TextEncoder + Unet + VAEDecoder + tokenizer). | |
| - Converted for on-device story / flashcard illustration in a German-learning app. | |
| ## Usage | |
| `StableDiffusionPipeline(resourcesAt: <snapshot>/original/compiled, ...)`. | |
| ## License | |
| Inherits **CreativeML OpenRAIL-M** from BK-SDM / Stable Diffusion v1.4 (see `LICENSE`). | |
| Commercial use and redistribution permitted; you must pass on the same use-restrictions. | |