Repository navigation
feat: add top-n-sigma sampler - #113
Merged
Merged
Conversation
Since nvcc is no longer wrapped with sccache, the long CUDA compile phase sends zero requests to the sccache server. With the default 600s idle timeout the server shuts itself down mid-build, so the trailing cl.exe objects fail with 'Timed out waiting for server startup' on the cold restart. Setting SCCACHE_IDLE_TIMEOUT=0 keeps the server alive for the whole build.
Expose llama.cpp's top-n-sigma sampler as a configurable advanced model setting. Keeps only tokens within n standard deviations of the top logit and is meant to replace top-k/top-p/min-p. Wired through session, model and global settings resolution, the sampler chain/order editor, provider parameter support info, feature generation and i18n (de/en).
Register a "qwen-image-2.1" stable-diffusion.cpp profile with Q3_K/Q4_K/Q6_K diffusion variants, Qwen3-VL-8B text/vision encoders and the Qwen-Image-2.1 VAE. Supports text-to-image and multi-reference editing (up to 10 refs), native transparency and strong in-image text. Wires up HF bundle detection (qwen-image prefix matching, qwen3-vl-8b arch hint), the LoRA architecture label and matcher rules. Also handle stable-diffusion.cpp master-962 (PR #2132) turning --auto-fit into a value-taking option: newer/unknown builds get `--auto-fit on`, older builds keep the bare flag.
Add the top-n-sigma sampler strings (label, description, short label and sampler-order entry) to the remaining 18 locales; previously only de/en were covered. Also drop the Greek sigma glyph from the short label in de/en in favour of the plain "Top-n-sigma".
Default sampling steps for the qwen-image-2.1 profile reduced from 40 to 35.
Collaborator
Author
|
Fixed the missing I18n-labels and added local Qwen Image 2.1 support |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The last commits where not merged and with 2.2.7 in the pipeline, this should be added.
This adds the (seemingly uncomitted?) fix for the desktop releases and also adds the top-n-sigma-sampler for local models.