model : Qwen3 Next (#16095)

* Qwen3 Next - cleaned up version

* Whitespaces and stuff

* Correct minor errors

* Update src/llama-model.cpp

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* Misc. fixes.

* Clean up code, add missing hybrid qualifier

* Did someone transpose the SOLVE_TRI result matrix? Perhaps...

* Whitespace

* Proper tensors for cb calls

* Use llama-graph.h vertical alignment

* BROKEN: chunking

* Set new tensors as inputs.

* Proper chunk logic

* It's the circle of life...

* More shenanigans for n_seq > 1

* Nail in the coffin?

* Fix Windows build

* Eh, one fails on Windows, the other fails on Mac... just use general capture.

* quant : cleanup

* model : cleanup

* qwen3 : cleanup

* cont : cleanup

* cont : cleanup

* ggml : revert change

* qwen3 : cleanup

* cont : cleanup

* Readd cmath

* qwen3 : fix typo

* Update convert_hf_to_gguf.py

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* Usual suspects

* fix my bad suggestion

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

src/models/models.h

@@ -2,8 +2,9 @@
 
 #include "../llama-model.h"
 #include "../llama-graph.h"
-#include "../llama-memory-recurrent.h"
 
+// TODO: remove in follow-up PR - move to .cpp files
+#include "../llama-memory-recurrent.h"
 #include <cmath>
 
 struct llm_graph_context_mamba : public llm_graph_context {
@@ -421,7 +422,56 @@ struct llm_build_qwen3vl : public llm_graph_context {
 struct llm_build_qwen3vlmoe : public llm_graph_context {
     llm_build_qwen3vlmoe(const llama_model & model, const llm_graph_params & params);
 };
+struct llm_build_qwen3next : public llm_graph_context_mamba {
+    llm_build_qwen3next(const llama_model & model, const llm_graph_params & params);
+private:
+    ggml_tensor * build_layer_attn(
+    llm_graph_input_attn_kv * inp_attn,
+                ggml_tensor * cur,
+                ggml_tensor * inp_pos,
+                        int   il);
 
+    ggml_tensor * build_layer_attn_linear(
+         llm_graph_input_rs * inp,
+                ggml_tensor * cur,
+                ggml_tensor * causal_mask,
+                ggml_tensor * identity,
+                        int   il);
+
+    ggml_tensor * build_layer_ffn(
+                ggml_tensor * cur,
+                        int   il);
+
+    ggml_tensor * build_delta_net_recurrent(
+                ggml_tensor * q,
+                ggml_tensor * k,
+                ggml_tensor * v,
+                ggml_tensor * g,
+                ggml_tensor * beta,
+                ggml_tensor * state,
+                ggml_tensor * causal_mask,
+                ggml_tensor * identity,
+                        int   il);
+
+    ggml_tensor * build_delta_net_chunking(
+                ggml_tensor * q,
+                ggml_tensor * k,
+                ggml_tensor * v,
+                ggml_tensor * g,
+                ggml_tensor * beta,
+                ggml_tensor * state,
+                ggml_tensor * causal_mask,
+                ggml_tensor * identity,
+                        int   il);
+
+    ggml_tensor * build_norm_gated(
+                ggml_tensor * input,
+                ggml_tensor * weights,
+                ggml_tensor * gate,
+                        int   layer);
+
+    const llama_model & model;
+};
 
 struct llm_build_qwen : public llm_graph_context {
     llm_build_qwen(const llama_model & model, const llm_graph_params & params);