* feat(devices): add choose_anima_inference_dtype honoring config.precision
When precision='auto', delegates to the existing choose_bfloat16_safe_dtype
probe. When set to a specific value (float16/bfloat16/float32), returns that
dtype directly so the user's config choice is honored. Anima call sites will
migrate to this helper in a follow-up commit.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat(anima): honor config.precision at the three Anima dtype call sites
Swaps choose_bfloat16_safe_dtype for choose_anima_inference_dtype at:
- _run_diffusion (transformer / latents / LoRA dtype)
- anima_text_encoder (text-encoder LoRA dtype)
- model_loaders/anima (transformer load dtype)
precision='auto' preserves the current bf16-probe fallback; specific values
are now honored end-to-end.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---------
Co-authored-by: Your Name <you@example.com>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Lincoln Stein <lincoln.stein@gmail.com>
Co-authored-by: Jonathan <34005131+JPPhoto@users.noreply.github.com>
kappacommit
61d1eaa5c5