refactor: Rename llm_build_hybrid_mamba -> llm_build_granite_hybrid

I've got back-and-forth a lot about how/if to try to implement reuse of the
"child model" layer types for hybrid models. At the end of the day, I think
hybrid models are their own beast and even if their layers are inspired by
other models, they should maintain control of their own layer building (in
other words, the copy-paste method). Given that, the name should reflect
that this is not a generic hybrid model builder, but rather a granite-
specific hybrid model builder that can do MoE (granite 4) or dense (bamba).

As part if this, I also cleaned up dangling comments from previous attempts
at using static methods for reusability.

Branch: GraniteFour

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

Author: gabe-l-hart

src/llama-model.cpp

@@ -13069,8 +13069,6 @@ struct llm_build_granite : public llm_graph_context {
         ggml_build_forward_expand(gf, cur);
     }
 
-    // static layer build function that enables other models to borrow this
-    // layer logic
     ggml_tensor * build_granite_attention_layer(
               ggml_cgraph                     * gf,
               ggml_tensor                     * cur,
@@ -13134,7 +13132,6 @@ struct llm_build_granite : public llm_graph_context {
         return cur;
     }
 
-    // static ffn layer builder for reuse in hybrid architectures
     ggml_tensor * build_layer_ffn(
               ggml_tensor       * cur,
               ggml_tensor       * inpSA,
@@ -13215,16 +13212,17 @@ struct llm_build_granite : public llm_graph_context {
     }
 };
 
-struct llm_build_hybrid_mamba : public llm_graph_context {
+struct llm_build_granite_hybrid : public llm_graph_context {
 
     const llama_model & model;
 
-    llm_build_hybrid_mamba(
-        const llama_model & model,
-        const llm_graph_params & params,
-        ggml_cgraph * gf,
-        const bool use_rope = true)
-    : llm_graph_context(params), model(model) {
+    llm_build_granite_hybrid(
+                 const llama_model & model,
+            const llm_graph_params & params,
+                       ggml_cgraph * gf,
+               const bool use_rope = true) :
+        llm_graph_context(params), model(model) {
+
         const int64_t n_embd_head = hparams.n_embd_head_v;
         GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
 
@@ -13424,8 +13422,6 @@ struct llm_build_hybrid_mamba : public llm_graph_context {
         return cur;
     }
 
-    // static layer build function that enables other models to borrow this
-    // layer logic
     ggml_tensor * build_granite_attention_layer(
               ggml_cgraph                * gf,
               ggml_tensor                * cur,
@@ -13489,7 +13485,6 @@ struct llm_build_hybrid_mamba : public llm_graph_context {
         return cur;
     }
 
-    // static ffn layer builder for reuse in hybrid architectures
     ggml_tensor * build_layer_ffn(
               ggml_tensor       * cur,
               ggml_tensor       * inpSA,
@@ -14859,12 +14854,12 @@ llm_graph_result_ptr llama_model::build_graph(
             } break;
         case LLM_ARCH_GRANITE_MOE_HYBRID:
             {
-                llm = std::make_unique<llm_build_hybrid_mamba>(*this, params, gf,
+                llm = std::make_unique<llm_build_granite_hybrid>(*this, params, gf,
                     /* use_rope   */ false);
             } break;
         case LLM_ARCH_BAMBA:
             {
-                llm = std::make_unique<llm_build_hybrid_mamba>(*this, params, gf,
+                llm = std::make_unique<llm_build_granite_hybrid>(*this, params, gf,
                     /* use_rope   */ true);
             } break;
         case LLM_ARCH_CHAMELEON: