[CPU] fix CPU backend sel. issue for Llama4 (#10511)

docs/platforms/cpu_server.md

@@ -81,6 +81,8 @@ git clone https://github.com/sgl-project/sglang.git
 cd sglang
 git checkout <YOUR-DESIRED-VERSION>
 
+# Use dedicated toml file
+cp python/pyproject_other.toml python/pyproject.toml
 # Install SGLang dependent libs, and build SGLang main package
 pip install --upgrade pip setuptools
 conda install -y libsqlite==3.48.0 gperftools tbb libnuma numactl

python/sglang/srt/server_args.py

@@ -2648,7 +2648,7 @@ class ServerArgs:
                 # use bf16 for mxfp4 triton kernels
                 self.dtype = "bfloat16"
 
-        elif "Llama4" in model_arch:
+        elif "Llama4" in model_arch and self.device != "cpu":
             assert self.attention_backend in {
                 "fa3",
                 "aiter",