Merge EmbeddedLLM/vllm-rocm into vLLM main (#1836)

Co-authored-by: Philipp Moritz <pcmoritz@gmail.com>
Co-authored-by: Amir Balwel <amoooori04@gmail.com>
Co-authored-by: root <kuanfu.liu@akirakan.com>
Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com>
Co-authored-by: kuanfu <kuanfu.liu@embeddedllm.com>
Co-authored-by: miloice <17350011+kliuae@users.noreply.github.com>

csrc/ops.h

@@ -61,12 +61,14 @@ void gelu_fast(
   torch::Tensor& out,
   torch::Tensor& input);
 
+#ifndef USE_ROCM
 torch::Tensor awq_gemm(
   torch::Tensor _in_feats,
   torch::Tensor _kernel,
   torch::Tensor _scaling_factors,
   torch::Tensor _zeros,
   int split_k_iters);
+#endif
 
 void squeezellm_gemm(
   torch::Tensor vec,