贡献 / #1108
#1108 Add architecture check condition in native_dense.cpp
open · @cdb0y511 · 0 评论 · 去 GitHub 看
说明
Fixes a split-GGUF loading issue reproduced with Huihui-Qwen3.8-Flash-Next-abliterated-GGUF (UD-Q4_K_XL). The first shard contains the full qwen4exp metadata, while later shards include general.architecture = qwen4exp but omit fields such as qwen4exp.block_count. Strata therefore re-validates the architecture on later shards and fails with: native dense projections: missing qwen4exp.block_count This change avoids repeating the full architecture check after it has already been successfully initialized from the first shard. The model loads and runs successfully after the fix.
本站相关内容
相关页面的快捷入口。