doc: document load_mode on the llm-llamacpp page instead of the removed no_mmap - #4115
Merged
Merged
Conversation
…ed no_mmap @qvac/llm-llamacpp 0.47.0 replaced the `no_mmap` load-config key with `load_mode`, but the page still lists `no_mmap` and describes it as a flag set by an empty string. Following it fails the load: the key is no longer consumed, so it reaches llama.cpp's argument parser as an unknown argument. The table now lists the five accepted values and the `mmap` default, checked against `kLoadModes` in the addon's `LoadFitNormalization.cpp` and the comment that absence applies llama.cpp's own default. `none` is described as a full buffered read, since it disables mlock and direct I/O as well as mmap. Note for readers of the example: `dio` is accepted by the addon but does not take effect in fabric v10297.0.0, so it is listed as a value without being recommended.
Contributor
Review StatusCurrent Status: ✅ APPROVED |
Contributor
License compliance — cleanNo new dependency license findings in this PR. Warn-only (shadow) mode — this check does not block merges yet. Updated automatically by the canonical license compliance workflow. NOTICE presence (advisory)Missing NOTICE (advisory, does not block):
|
BrunoCampana
approved these changes
Sep 2, 2026
BrunoCampana
approved these changes
Sep 2, 2026
iancris
approved these changes
Sep 2, 2026
gianni-cor
approved these changes
Sep 2, 2026
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
4 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
🎯 What problem does this PR solve?
@qvac/llm-llamacpp@0.47.0replaced theno_mmapload-config key withload_mode, but the addon page still listsno_mmapand describes it as a flag set by passing an empty string.invalid argument: --no-mmap.📝 How does it solve it?
no_mmapparameter row withload_mode, listing the five accepted values and themmapdefault — verified againstkLoadModesin the addon'saddon/src/model-interface/LoadFitNormalization.cppand the comment that absence applies llama.cpp's own default.load_mode: 'none'to the config example, described as a full buffered read: it disables mlock and direct I/O as well as mmap, which "disable memory mapping" did not convey.Docs only. Split out of the SDK change set for QVAC-24073 (#4078) to follow how this page is normally maintained — #3854 exposed
image_no_upscalein the same config schema with no docs, and #3959 updated the page separately under its own ticket.Worth knowing for readers:
diois accepted by the addon but does not take effect in the pinned fabricv10297.0.0— the loader setsuse_direct_io, whilegguf_file_loadconstructsllama_file_disk(path, "rb")without passing it. It is listed as an accepted value without being recommended; the native fix is tracked separately.