kv-cache : support V-less cache (#19067)

* kv-cache : support V-less cache

* cuda : better check for V_is_K_view

* cuda : improve V_is_K_view check

* graph : add comments

* hparams : refactor

ggml/src/ggml-cuda/fattn.cu

@@ -247,7 +247,7 @@ static best_fattn_kernel ggml_cuda_get_best_fattn_kernel(const int device, const
         }
     }
 
-    const bool V_is_K_view = V->op == GGML_OP_VIEW && V->src[0] == K && V->data == K->data;
+    const bool V_is_K_view = V->view_src && V->view_offs == 0 && (V->view_src == K || V->view_src == K->view_src);
 
     const int cc = ggml_cuda_info().devices[device].cc;