Merge branch 'add-fh1-rebased' of https://github.com/tiiuae/llama.cpp-public into add-fh1-rebased

Author: ibrahimkhadraoui

src/llama-model.cpp

@@ -4530,7 +4530,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                     // output
                     {
                         output = create_tensor(tn(LLM_TENSOR_OUTPUT, "weight"), {hidden_size, n_vocab}, TENSOR_NOT_REQUIRED);
-                        final_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {hidden_size}, 0);
+                        output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {hidden_size}, 0);
                     }
 
                     for (int i = 0; i < n_layer; ++i) {
@@ -14740,7 +14740,7 @@ struct llm_build_falcon_h1 : public llm_graph_context {
         cur = inpL;
 
         cur = build_norm(cur,
-                model.final_norm, NULL,
+                model.output_norm, NULL,
                 LLM_NORM_RMS, -1);
 
         cb(cur, "result_norm", -1);

src/llama-model.h

@@ -173,7 +173,6 @@ struct llama_layer {
     struct ggml_tensor * attn_norm_cross = nullptr;
     struct ggml_tensor * attn_norm_enc   = nullptr;
     struct ggml_tensor * ssm_norm        = nullptr;
-    struct ggml_tensor * final_norm      = nullptr;
 
     // attention
     struct ggml_tensor * wq        = nullptr;
@@ -365,7 +364,6 @@ struct llama_model {
     struct ggml_tensor * output          = nullptr;
     struct ggml_tensor * output_b        = nullptr;
     struct ggml_tensor * output_norm_enc = nullptr;
-    struct ggml_tensor * final_norm      = nullptr;
 
     // classifier
     struct ggml_tensor * cls       = nullptr;