llama.cpp Error: Ling 3.0 Flash Model Shows 'Unused Tensor' Warnings

Debianreiser69 · reddit · 2026-08-22

A user encountered errors while loading the Ling-3.0-flash-Q6K model (sharded GGUF) in llama.cpp. Logs show multiple tensors in blk.42 (e.g., attnq.weight, ffngateexps.weight) are marked as 'unused' and ignored. The user reports similar errors with other models like Tencent's Hunyuan. No solution is provided yet.

Original post →

More from Infra

Infra channel →