v0.1.6

0.1.6 (2026-06-13)

🐛 Bug Fixes

• deps: drop the stale <3.13 Python cap (allow >=3.12) (9b44a5a)

v0.1.5

0.1.5 (2026-06-13)

✨ Features

• convert: auto-detect model version from the checkpoint (2a24d4e)

v0.1.4

0.1.4 (2026-06-13)

🐛 Bug Fixes

• convert: generic LCM conversion for arbitrary checkpoints (bedeb49)
• inference: add output_dtype to CoreMLTextEncoder (eb6d3b5)

v0.1.3

0.1.3 (2026-06-04)

✨ Features

• convert: add VAE and CLIP text-encoder conversion (dc1f85b)
• inference: end-to-end Core ML pipeline (VAE + text-encoder swap) (ca08b16)

🐛 Bug Fixes

• inference: expose .device on the Core ML adapters (30a673e)

📚 Documentation

• readme: link the log.aszc.dev energy benchmark writeup (77927b5)

v0.1.2

0.1.2 (2026-05-27)

🐛 Bug Fixes

• attention: convertible fp32 ORIGINAL attention for the Core ML GPU path (#2) (28e56fc)

v0.1.1

fix(deps): drop torch upper bound to avoid host downgrade

The hard torch<2.8 cap forced uv to downgrade the host environment's torch on install (e.g. ComfyUI 2.12 -> 2.7.1), leaving torchvision/torchaudio compiled against the original torch and breaking their ABI (operator torchvision::nms does not exist). torch is supplied by the host; inference does not require the cap. coremltools 9 conversion remains validated against torch 2.7.x.

v0.1.0

Initial PyPI release of coreml-diffusion: framework-free Core ML diffusion conversion (SD1.5/SDXL) for Apple Neural Engine.