Commit Graph

191 Commits

Author SHA1 Message Date
Concedo 7f485e5287 remove CLBlast, part 1 2026-01-23 13:50:12 +08:00
Concedo cdd6578a9a esrgan added 2026-01-20 22:10:37 +08:00
Concedo c9c15749e0 wip on adding esrgan upscaling 2026-01-20 00:35:35 +08:00
Llama 95ebfdcde8 Add token ids to logprob data returned by the API (#1928)
Previously, logprobs only contained the token string
and byte data, as well as the log probability itself.
For workflows that require the token id, translating
from the token bytes to the token id is potentially
costly and unreliable. It is simple and inexpensive
to expose the numeric token ids directly instead.
2026-01-18 16:30:46 +08:00
Concedo 3816391a74 increase logprobs returned to 10 2026-01-18 11:13:42 +08:00
Concedo 62bea5ef4f allow overriding the devices directly 2026-01-17 19:08:06 +08:00
Concedo d8942cde14 smartcache allow custom number of slots 2026-01-02 17:19:40 +08:00
Concedo 442fa7cd7c support for circular textures in sdcpp 2026-01-01 16:34:09 +08:00
Concedo 27261bfc26 adaptive decay as an overridable param (+1 squashed commits)
Squashed commits:

[d94df7843] adaptive decay as an overridable param
2025-12-28 13:34:20 +08:00
Concedo 6548645aaa rename power law sampler to adaptive p 2025-12-27 17:50:58 +08:00
Concedo 91d8863f18 power law sampler added 2025-12-27 09:46:06 +08:00
Wagner Bruna f30da43b7f sd: get the available schedulers directly from sd.cpp (#1900)
Avoids a hardcoded list on the Python side.
2025-12-24 21:55:24 +08:00
Concedo 26d89bf589 support for downloading AVI from sdui 2025-12-24 18:40:10 +08:00
Concedo fedd529fdc autofit counts overheads 2025-12-21 14:31:08 +08:00
Concedo 1e083d9c8b integrate autofit for upstream, removed forceversion 2025-12-17 18:42:47 +08:00
Concedo ab9bc6f2ae zimage cfg clamp is opt out with remove_limits 2025-12-13 09:20:00 +08:00
Concedo fd0d0cab03 move pipeline parallelism to a --pipelineparallel launch flag 2025-12-11 21:03:41 +08:00
Concedo 8a18e094f5 added smartcaching implementation inspired from Pento95 (+2 squashed commit)
Squashed commit:

[fcc498688] wip basic smart caching test

[b6e8b2577] wip basic smart caching test
2025-12-10 18:00:03 +08:00
Wagner Bruna 3a7dd1a97f sd: sync to master-358-347710f
Also adapt Koboldcpp LoRA loading function, and add
backend support for lora_apply_mode.
2025-11-23 19:28:54 -03:00
LostRuins Concedo 281542aa0d add smoothing curve, not tested 2025-11-17 23:07:35 +08:00
LostRuins Concedo a67f097de3 fixed incorrect parameter order for mirostat 2025-11-16 17:38:25 +08:00
Concedo 57e1d9c822 rename blasbatchsize to batchsize 2025-10-24 18:16:54 +08:00
Concedo 68c9d955d2 support multiple override kv 2025-10-24 17:28:54 +08:00
Wagner Bruna c48999f7c0 additional options for image generation (#1765)
* sd: add backend support for choosing the default sampler

* use the default sampler on the API

* sd: add backend support for the scheduler

* sd: add backend support for distilled guidance

* sd: add backend support for timestep-shift

* sd: add a config field to set default image gen options
2025-10-05 23:36:20 +08:00
Concedo f282362414 added qwen image support (+1 squashed commits)
Squashed commits:

[92df28061] added qwen image support (+1 squashed commits)

Squashed commits:

[1485c71ed] wip adding qwen image
2025-10-03 18:58:48 +08:00
Concedo 2fc31d36c0 gif mime type for animated images 2025-10-01 17:18:00 +08:00
Concedo 4b1c89ca5c can save animated gifs 2025-09-29 22:52:42 +08:00
Concedo 7b71742223 added additional toggles for SD, sdoffloadcpu , sdvaecpu and sdclipcpu 2025-09-28 22:08:29 +08:00
Concedo 553890f975 compact UI launcher, WIP handling video outputs 2025-09-28 17:24:04 +08:00
Concedo 3210b378e8 better tool calls 2025-08-20 22:11:31 +08:00
Wagner Bruna 6003e90e50 Add flash attention and conv2d direct controls for image generation (#1678)
* Add separate flash attention config for image generation

* Add config option for Conv2D Direct
2025-08-20 12:17:57 +08:00
Concedo 2abe11071b custom voice handling 2025-08-18 16:57:34 +08:00
Concedo 5a921a40f9 add overridenativecontext flag, stop nagging me 2025-08-14 22:54:45 +08:00
Concedo e40d26b9e7 allow offloading moe to cpu with --moecpu 2025-08-05 23:42:42 +08:00
Concedo abf527a207 clearer multimodal capability display 2025-07-28 22:54:49 +08:00
Concedo 811463a704 split audio and vision detection separately 2025-07-13 17:47:15 +08:00
Concedo e9473305d0 wip2 (+1 squashed commits)
Squashed commits:

[4628777b6] wip
2025-07-12 18:54:40 +08:00
Wagner Bruna d74c16e6e0 enable flash attention for image generation (#1633) 2025-07-05 11:20:51 +08:00
Concedo 4ec0e0fd21 now accept multiple images for reference images 2025-06-28 17:30:28 +08:00
Concedo 0bd648ffa4 photomaker renamed to extra image to handle future extension 2025-06-28 10:26:06 +08:00
Concedo 65ff041827 added more perf stats 2025-06-21 12:12:28 +08:00
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
Concedo 4e40f2aaf4 added photomaker face cloning 2025-06-20 21:33:36 +08:00
Concedo 21881a861d rename restrict square to sdclampedsoft 2025-06-20 15:39:55 +08:00
Wagner Bruna f6d2d1ce5c configurable resolution limit (#1586)
* refactor image gen configuration screen

* make image size limit configurable

* fix resolution limits and keep dimensions closer to the original ratio

* use 0.0 for the configured default image size limit

This prevents the current default value from being saved into the
config files, in case we later decide to adopt a different value.

* export image model version when loading

* restore model-specific default image size limit

* change the image area restriction to be specified by a square side

* move image resolution limits down to the C++ level

* Revert "export image model version when loading"

This reverts commit fa65b23de3.

* Linting Fixes:
PY:
- Inconsistent var name sd_restrict_square -> sd_restrict_square_var
- GUI swap back to using absolute row numbers for now.
- fstring fix
- size_limit -> side_limit inconsistency
C++:
- roundup_64 standalone function
- refactor sd_fix_resolution variable names for clarity
- move "anti crashing" hard total megapixel limit always to be applied after soft total megapixel limit instead of conditionally only when sd_restrict_square is unset

* allow unsafe resolutions if debugmode is on

---------

Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com>
2025-06-13 20:05:20 +08:00
Concedo 1cbe716e45 allow setting maingpu 2025-06-12 17:53:43 +08:00
Concedo 7d8aa31f1f fixed embeddings, added new parameter to limit max embeddings context 2025-06-10 01:11:55 +08:00
Concedo cfcdfd69bd allow embeddings models to use mmap 2025-06-07 10:14:00 +08:00
Concedo 6ce85c54d6 not working correctly 2025-06-02 22:12:10 +08:00
Concedo 8e1ebc55b5 dropped support for lora base as upstream no longer uses it. If provided it will be silently ignored 2025-06-02 12:49:53 +08:00