cuda: raise CUDA graph cache cap from 64 to 512 for tensor parallel - #232

cuda: raise CUDA graph cache cap from 64 to 512 for tensor parallel#232

angliubc wants to merge 1 commit into

unslothai:mtp/qwen4exp-nextnunslothai/llama.cpp:mtp/qwen4exp-nextnfrom

angliubc:fix/cuda-graph-cache-cap-tpangliubc/llama.cpp:fix/cuda-graph-cache-cap-tpCopy head branch name to clipboard