models : nomic bert moe does not require KV cache

Author: ggerganov

src/llama-model.cpp

@@ -12836,6 +12836,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
         case LLM_ARCH_BERT:
         case LLM_ARCH_JINA_BERT_V2:
         case LLM_ARCH_NOMIC_BERT:
+        case LLM_ARCH_NOMIC_BERT_MOE:
             {
                 res = nullptr;
             } break;