Repository navigation
Publish the confirmed Nebula engine and English release package - #1
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The repository now contains the confirmed Q4 Flash-Next engine: fixed GPU experts, CPU MoE on handoff, retained-prefix recovery, time-aware MTP and CUDA graphs. The English README follows the approved presentation and includes its logo, component tables, hardware profiles and completed standard benchmarks.
The package adds an automatic Windows/WSL2 installer, the pinned tokenizer, complete expert ranking, 16 hardware profiles, SSD streaming, separate model license and GitHub checks. Obsolete experimental preparation scripts, the light-model path, the Italian README and the machine-specific launcher are removed.
Validation: 105 source/data files checked in the GitHub tree; all 16 native C/CUDA/header files match the confirmed source; 89 Windows assertions; 39 Python application tests plus three preparation tests; native RAM/SSD, handoff and scalar/AVX2/AVX-512 equivalence checks. Two CUDA-only tests are excluded from the CPU packaging run. The rebuilt setup payload has 72 checksummed files. Benchmark evidence contains 140 original saved answers and 70 pairs, with no generation rerun.
The setup executable is a release candidate pending a complete clean-machine unattended installation. Repository visibility remains private.