Groundwork for image-grounded conditioning (rung B). Vendors mlx-vlm's
pure-MLX Qwen3-VL vision tower into krea2_engine/vision/ (self-contained:
inlined SDPA helper, local VisionConfig) and adds torch-free image
preprocessing (smart-resize + CLIP normalize + patch flatten).
Both validated against the PyTorch reference (tools/test_vision_tower.py):
- tower vs transformers Qwen3VLVisionModel: merged embeds + all 3 deepstack
features match to max|Δ|~2e-4, cosine 1.0 (bf16 rounding).
- preprocessing vs transformers image processor: max|Δ|~2e-7, grid exact.
Loads the visual.* weights already present in the Krea-2-Turbo text
encoder (315 tensors, 0 missing). Not yet wired into any node — the
grounded-encode integration (M-RoPE, deepstack injection, 12-layer tap)
is the next step. mlx-vlm is MIT; attribution added to NOTICE.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
README documents the measured speedups (512² ~35% / 1024² ~15% on
M1 Max), the prompt-embedding cache, and the KREA2_EXACT_LEGACY seed-
reproducibility flag. NOTICE records the performance modifications to
the vendored engine, as its attribution chain previously stated the
engine was otherwise unmodified.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Krea2ModelLoader / Krea2LoRA / Krea2Generate / Krea2Unload nodes running
Krea-2-Turbo on Apple MLX in-process, returning ComfyUI IMAGE tensors.
The krea2_engine/ pipeline is vendored directly in this repo (no external
git dependency) — see NOTICE for the full attribution chain back to the
original Krea 2 model and the MLX port it builds on.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
