Commit 85ca3b52c for llama.cpp

commit 85ca3b52c33f5477f75985a251543f5f73010e8a
Author: Georgi Gerganov <ggerganov@gmail.com>
Date:   Sun Sep 27 10:16:23 2026 +0300

    hrm : fix layer placement of `z_l_init` weight (#29512)

diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp
index 9d2331c2f..8f1e239da 100644
--- a/src/llama-arch.cpp
+++ b/src/llama-arch.cpp
@@ -722,7 +722,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
     {LLM_TENSOR_TOKEN_EMBD,                 {LLM_TENSOR_LAYER_INPUT,     GGML_OP_GET_ROWS}},
     {LLM_TENSOR_POS_EMBD,                   {LLM_TENSOR_LAYER_INPUT,     GGML_OP_GET_ROWS}},
     {LLM_TENSOR_TOKEN_TYPES,                {LLM_TENSOR_LAYER_INPUT,     GGML_OP_GET_ROWS}},
-    {LLM_TENSOR_HRM_Z_L_INIT,               {LLM_TENSOR_LAYER_INPUT,     GGML_OP_ADD}},
+    {LLM_TENSOR_HRM_Z_L_INIT,               {LLM_TENSOR_LAYER_REPEATING, GGML_OP_ADD}},
     {LLM_TENSOR_TOKEN_EMBD_NORM,            {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},  // do the norms on the first layer (not the input layer)
     {LLM_TENSOR_OUTPUT,                     {LLM_TENSOR_LAYER_OUTPUT,    GGML_OP_MUL_MAT}},
     {LLM_TENSOR_CLS,                        {LLM_TENSOR_LAYER_OUTPUT,    GGML_OP_MUL_MAT}},
diff --git a/src/models/hrm-text.cpp b/src/models/hrm-text.cpp
index 4a9a67b6c..8b3337c0b 100644
--- a/src/models/hrm-text.cpp
+++ b/src/models/hrm-text.cpp
@@ -43,7 +43,7 @@ void llama_model_hrm_text::load_arch_tensors(llama_model_loader &) {
         output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }, TENSOR_DUPLICATED);
     }

-    hrm_z_l_init = create_tensor(tn(LLM_TENSOR_HRM_Z_L_INIT), { n_embd }, 0);
+    hrm_z_l_init = create_tensor(tn(LLM_TENSOR_HRM_Z_L_INIT, 0), { n_embd }, 0);

     const int lps = hparams.n_hrm_layers_per_stack;