returnmoe

/

jlens-adapters

like 3

Follow

return moe 2

Documents the exact Qwen/Qwen3.6-27B revision recovered from the original fitting cache.

Evidence

• Cache root used by the fitting jobs: /scratch/u710878/miru-qwen36-27b/hf
• Hugging Face cache refs/main: 6a9e13bd6fc8f0983b9b99948120bc37f49c13e9
• The cache contains exactly one model snapshot, with the same SHA.
• The cached ref and model files predate the fitted adapter: 2026-07-16 versus 2026-07-21.
• The revision resolves successfully in the upstream Qwen/Qwen3.6-27B repository.

The revision was reconstructed afterward from the preserved cache, then written into the adapter exactly where the fitter records it: fit_metadata.provenance.model_commit_hash.

• Revised adapter SHA-256: c1ebc7b2e1dee5b16cf308e2655b261ebba7744ed52212ca2551870d969011a5
• Adapter structure after rewriting: 63 float16 matrices, each 5120 x 5120
• Embedded model_commit_hash: 6a9e13bd6fc8f0983b9b99948120bc37f49c13e9

Contributor Public-Domain Certification

I certify that I created this contribution or otherwise have the authority
to submit it. To the extent that I own copyright or related rights in the
adapter and its accompanying metadata, I permanently dedicate those rights to
the public domain under the Unlicense. Where a public-domain dedication is
not legally recognized, I make the contribution available under all
permissions and disclaimers stated by the Unlicense. I have disclosed the
base model and calibration sources, and I am not knowingly submitting
material that I lack permission to distribute. If I am contributing as part
of my employment or for another organization, I certify that I am authorized
to make this dedication on its behalf.

Signed-off-by: Lucas Teske (@racerxdl ), 2026-07-26