Move the TurboQuant backend index to unified release tags - #10
Merged
Conversation
The fork now publishes every backend variant of a build under one release tag (b<upstream-build>-<fork-semver>, e.g. b10018-1.3.0) instead of one release per variant, and that release adds Linux CPU, CUDA 12.4/13.3 and ROCm artifacts. Pin the index to that release, list the new Linux tiers plus the macOS entry the build system resolves, and replace the per-variant tag invariant with a unified-tag + per-platform-extension contract. Co-authored-by: Cursor <cursoragent@cursor.com>
The Windows CUDA archives of b10018-1.3.0 ship ggml-cuda.dll but no CUDA runtime DLLs, so the client's cudart companion download is still required. Verified by unpacking the published archive. Co-authored-by: Cursor <cursoragent@cursor.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
backends/turboquant-manifest.jsonto the unifiedb<upstream-build>-<fork-semver>release model. Every entry now shares one tag,b10018-1.3.0, instead of a per-variantturboquant-<id>-<sha>tag.linux-x64-cpu,linux-x64-cuda-12.4,linux-x64-cuda-13.3andlinux-x64-rocm.linux-x64-vulkanstays the tier the installer bundles as the offline fallback.turboquant-schema.jsonand the CI integrity check: the tag must match^b\d+-\d+\.\d+\.\d+$, all entries must share one tag, and the asset extension is enforced per platform (.zipon Windows,.tar.gzelsewhere).ggml-cuda.dllbut no CUDA runtime DLLs, so the client'scudart-*companion download is still required.Test plan
ajvvalidates the manifest against the schemab10018-1.3.0.zipduplicates of the Linux/macOS archives are deliberately not listedwindows-x64-cpuandwindows-x64-vulkanfrom the release: layout is stillbuild/bin/llama-server.exe, the binary reportsversion: b10018-1.3.0 (build 10408, commit 5bc5c248d), matching the manifestcommitfield, and the Vulkan tier enumerates both GPUsMade with Cursor