diff --git a/llama.cpp b/llama.cpp
index d840d955e..d1e72c9bc 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -4529,7 +4529,14 @@ static bool llm_load_tensors(
                     {
                         model.output_norm = ml.create_tensor(ctx_output,       tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
                         model.output_norm_b = ml.create_tensor(ctx_output,     tn(LLM_TENSOR_OUTPUT_NORM, "bias"),   {n_embd});
-                        model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab});
+                        
+                        model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, false);
+                        // if output is NULL, init from the input tok embed
+                        if (model.output == NULL) {
+                            model.output = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
+                            ml.n_created--; // artificial tensor
+                        }
+
                     }
 
                     for (int i = 0; i < n_layer; ++i) {
@@ -4560,7 +4567,7 @@ static bool llm_load_tensors(
 
                         // optional bias tensors
                         layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, false);
-                        layer.ffn_up_b   = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP , "bias", i), { n_embd}, false);
+                        layer.ffn_up_b   = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP , "bias", i), { n_ff}, false);
                     }
                 } break;
             default: