fix: resolve LLM norm tensor names by architecture #2453

Job	Run time
windows-latest-rocm	6m 5s
macOS-latest-cmake	8m 5s
ubuntu-latest-cmake	5m 17s
ubuntu-latest-rocm (7.2.1, gfx908;gfx90a;gfx942;gfx1030;gfx1031;gfx1032;gfx1100;gfx1101;gfx1102;g...	5m 16s
ubuntu-latest-cmake-vulkan	9m 45s
windows-latest-cmake-hip	4m 27s
windows-latest-cmake (avx2, -DGGML_NATIVE=OFF -DGGML_AVX2=ON -DSD_BUILD_SHARED_LIBS=ON)	3m 40s
ubuntu-latest-rocm (7.13.0, gfx906;gfx908;gfx90a;gfx942;gfx950;gfx1010;gfx1011;gfx1012;gfx1030;gf...	5m 28s
windows-latest-cmake (avx512, -DGGML_NATIVE=OFF -DGGML_AVX512=ON -DGGML_AVX=ON -DGGML_AVX2=ON -DS...	4m 10s
windows-latest-cmake (vulkan, -DSD_VULKAN=ON -DSD_BUILD_SHARED_LIBS=ON)	7m 53s
windows-latest-cmake (avx, -DGGML_NATIVE=OFF -DGGML_AVX=ON -DGGML_AVX2=OFF -DSD_BUILD_SHARED_LIBS...	4m 16s
windows-latest-cmake (noavx, -DGGML_NATIVE=OFF -DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF -DSD...	3m 40s
windows-latest-cmake (cuda12, -DSD_CUDA=ON -DSD_BUILD_SHARED_LIBS=ON -DCMAKE_CUDA_ARCHITECTURES='...	2h 4m 1s
Build and push container images	0s
release	-1s
	3h 12m 2s

Provide feedback