Files
koboldcpp/otherarch
Wagner Bruna 08adfb53c9 Configurable VAE threshold limit (#1601)
* add backend support for changing the VAE tiling threshold

* trigger VAE tiling by image area instead of dimensions

I've tested with GGML_VULKAN_MEMORY_DEBUG all resolutions with
the same 768x768 area (even extremes like 64x9216), and many
below that: all consistently allocate 6656 bytes per image pixel.
As tiling is primarily useful to avoid excessive memory usage, it
seems reasonable to enable VAE tiling based on area rather than
maximum image side.

However, as there is currently no user interface option to change
it back to a lower value, it's best to maintain the default
behavior for now.

* replace the notile option with a configurable threshold

This allows selecting a lower threshold value, reducing the
peak memory usage.

The legacy sdnotile parameter gets automatically converted to
the new parameter, if it's the only one supplied.

* simplify tiling checks, 768 default visible in launcher

---------

Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com>
2025-06-21 10:14:57 +08:00
..
2025-06-21 08:40:04 +08:00
2025-03-12 18:58:27 +08:00
2025-06-13 22:21:15 +08:00
2023-09-16 18:52:09 +08:00
2023-05-20 15:58:33 +08:00
2024-09-19 13:56:19 +08:00
2025-05-21 21:47:40 +08:00
2023-05-24 00:20:43 +08:00
2023-09-16 18:52:09 +08:00
2023-09-16 18:52:09 +08:00
2025-01-13 23:26:54 +08:00
2025-01-13 23:26:54 +08:00
2023-05-24 00:20:43 +08:00
2023-05-27 22:45:28 +08:00
2024-08-25 23:15:38 +08:00
2023-07-10 18:23:25 +08:00
2024-05-14 19:17:01 +08:00
2025-06-13 22:21:15 +08:00
2025-04-14 23:05:36 +08:00
2025-04-14 23:05:36 +08:00