CUDA: fix logic for clearing padding with -ngl 0#13320

JohannesGaessler merged 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

JohannesGaessler:cuda-fix-deepseek-partial-offloadJohannesGaessler/llama.cpp:cuda-fix-deepseek-partial-offloadCopy head branch name to clipboard