llama : add assert about missing llama_encode() call (ggml-org#8400)

Co-authored-by: Stanisław Szymczyk <[email protected]>

Author: 2 people

src/llama.cpp

@@ -13201,6 +13201,8 @@ struct llm_build_context {
                     LLM_NORM_RMS, cb, -1);
             cb(cur, "result_norm", -1);
         } else {
+            GGML_ASSERT(n_outputs_enc > 0 && "call llama_encode() first");
+
             struct ggml_tensor * embd_enc       = llm_build_inp_embd_enc();
             struct ggml_tensor * pos_bucket_dec = llm_build_pos_bucket(true);