llama : de-duplicate deepseek2 norm

Author: ggerganov

src/llama.cpp

@@ -12873,12 +12873,12 @@ struct llm_build_context {
             struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
             cb(ffn_inp, "ffn_inp", il);
 
-            if ((uint32_t) il < hparams.n_layer_dense_lead) {
-                cur = llm_build_norm(ctx0, ffn_inp, hparams,
-                        model.layers[il].ffn_norm, NULL,
-                        LLM_NORM_RMS, cb, il);
-                cb(cur, "ffn_norm", il);
+            cur = llm_build_norm(ctx0, ffn_inp, hparams,
+                    model.layers[il].ffn_norm, NULL,
+                    LLM_NORM_RMS, cb, il);
+            cb(cur, "ffn_norm", il);
 
+            if ((uint32_t) il < hparams.n_layer_dense_lead) {
                 cur = llm_build_ffn(ctx0, cur,
                         model.layers[il].ffn_up,   NULL, NULL,
                         model.layers[il].ffn_gate, NULL, NULL,
@@ -12888,11 +12888,6 @@ struct llm_build_context {
                 cb(cur, "ffn_out", il);
             } else {
                 // MoE branch
-                cur = llm_build_norm(ctx0, ffn_inp, hparams,
-                        model.layers[il].ffn_norm, NULL,
-                        LLM_NORM_RMS, cb, il);
-                cb(cur, "ffn_norm", il);
-
                 ggml_tensor * moe_out =
                         llm_build_moe_ffn(ctx0, cur,
                             model.layers[il].ffn_gate_inp,