Commit History
13 days ago
-
docs: sync model card with NInfer master
Verified signature0190293
-
docs: refresh FP8 KV serving performance
Verified signatureSync the September 28-29 Qwen3.8-27B NVFP4 serving measurements and their conditions and limitations from NInfer commit d23835c1.a107ba1
-
Update the built-in Qwen chat template
Verified signaturef0b43ad -
Publish v3 artifact and updated model card
Verified signature6a534f2
-
docs: sync model card with ninfer
Verified signatureSynchronize the version-controlled model card from https://github.com/Neroued/ninfer/commit/a16b6442856620b7e4856acb25215acbf7e3c750.11dbbbb
-
Update artifact with DFlash2 companion weights
Verified signatureIncludes the five-layer DFlash2 draft and selector weights. Update size, SHA-256, inventory, and runtime requirements together.5290713
-
docs(eval): add official qwen3.8 bf16 comparison to model card
Verified signature204e3d9 -
docs(eval): publish qwen3.8 nvfp4 reasoning results
Verified signature7940261
-
Update GPQA-Diamond evaluation
Verified signature7ebc71d
-
Publish Qwen3.8 NVFP4 MTP0 performance
Verified signatured6d0b3b -
Publish Qwen3.8 NVFP4 serving performance
Verified signatureAdd MTP3 single-request and concurrent corpus-makespan results.19415e1 -
Publish NInfer Qwen3.8-27B NVFP4 artifact
Verified signatureRegistered qwen3.8-27b/nvfp4 version-2 artifact, model card, manifest, checksum, and license.3b84117
-
initial commit
Verified signaturec3b02cf