Files
ik_llama.cpp/examples/llama-bench/llama-bench.cpp
Kawrakow d239dabcc6 Graph parallel for Qwen-3.5-MoE (#1347)
* Graph parallel for Qwen3.5-MoE

* Add --max-gpu to llama-bench

* Fix graph reuse when not all GPUs participate in self-attention
2026-03-02 07:48:43 +01:00

80 KiB