From 59bf83ccd245ef7e97a3a7576406ab8a77adbd17 Mon Sep 17 00:00:00 2001
From: Georgi Gerganov <ggerganov@gmail.com>
Date: Thu, 6 Aug 2026 08:16:15 +0300
Subject: [PATCH] wip

---
 src/llama-graph.cpp | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/llama-graph.cpp b/src/llama-graph.cpp
index fd4290cf05..59b52abd22 100644
--- a/src/llama-graph.cpp
+++ b/src/llama-graph.cpp
@@ -2451,6 +2451,7 @@ ggml_tensor * llm_graph_context::build_inp_pos() const {
 
     cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, (int64_t)n_tokens*hparams.n_pos_per_embd());
     ggml_set_input(cur);
+    cb(cur, "inp_pos", -1);
 
     res->add_input(std::move(inp));
 
@@ -2465,7 +2466,7 @@ ggml_tensor * llm_graph_context::build_inp_attn_scale() const {
     // this need to be 1x1xN for broadcasting
     cur = ggml_new_tensor_3d(ctx0, GGML_TYPE_F32, 1, 1, n_tokens);
     ggml_set_input(cur);
-    ggml_set_name(cur, "attn_scale");
+    cb(cur, "inp_attn_scale", -1);
 
     res->add_input(std::move(inp));
 
@@ -2487,6 +2488,7 @@ ggml_tensor * llm_graph_context::build_inp_out_ids() const {
 
     cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_outputs);
     ggml_set_input(cur);
+    ggml_set_name(cur, "out_ids");
 
     res->add_input(std::move(inp));
 
@@ -2500,6 +2502,7 @@ ggml_tensor * llm_graph_context::build_inp_mean() const {
 
     cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_tokens, ubatch.n_seqs_unq);
     ggml_set_input(cur);
+    ggml_set_name(cur, "mean");
 
     res->add_input(std::move(inp));
 
@@ -2513,6 +2516,7 @@ ggml_tensor * llm_graph_context::build_inp_cls() const {
 
     cur = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, ubatch.n_seqs_unq);
     ggml_set_input(cur);
+    ggml_set_name(cur, "cls");
 
     res->add_input(std::move(inp));
 
@@ -2537,6 +2541,7 @@ ggml_tensor * llm_graph_context::build_inp_cross_embd() const {
 
     cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, n_enc);
     ggml_set_input(cur);
+    ggml_set_name(cur, "cross_embd");
 
     res->add_input(std::move(inp));
 
@@ -2550,6 +2555,7 @@ ggml_tensor * llm_graph_context::build_inp_pos_bucket_enc() const {
 
     cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_tokens, n_tokens);
     ggml_set_input(cur);
+    ggml_set_name(cur, "pos_bucket_enc");
 
     res->add_input(std::move(inp));
 
@@ -2567,6 +2573,7 @@ ggml_tensor * llm_graph_context::build_inp_pos_bucket_dec() const {
 
     cur = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_kv, n_tokens);
     ggml_set_input(cur);
+    ggml_set_name(cur, "pos_bucket_dec");
 
     res->add_input(std::move(inp));
 
@@ -2735,6 +2742,7 @@ llm_graph_input_attn_no_cache * llm_graph_context::build_attn_inp_no_cache() con
     // note: there is no KV cache, so the number of KV values is equal to the number of tokens in the batch
     inp->self_kq_mask = ggml_new_tensor_4d(ctx0, type_mask, n_tokens, n_tokens, 1, 1);
     ggml_set_input(inp->self_kq_mask);
+    cb(inp->self_kq_mask, "self_kq_mask", -1);
 
     inp->self_kq_mask_cnv = inp->self_kq_mask;