Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

ggml: add GATED_DELTA_NET op (#19504)

Browse Source 
* ggml: add GATED_DELTA_NET op

* remove the transpose

* add KDA

* add qwen35 dense

* llama : check for fused gated delta net backend support

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This commit is contained in:
Aman Gupta
2026-03-07 15:41:10 +08:00
committed by GitHub
co-authored by Georgi Gerganov
parent 6fce5c6a7d
commit c5a778891b
15 changed files with 627 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-cpu/ggml-cpu.c
+10
View File
@@ -2021,6 +2021,10 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
{
ggml_compute_forward_solve_tri(params, tensor);
} break;
case GGML_OP_GATED_DELTA_NET:
{
ggml_compute_forward_gated_delta_net(params, tensor);
} break;
case GGML_OP_MAP_CUSTOM1:
{
ggml_compute_forward_map_custom1(params, tensor);
@@ -2200,6 +2204,7 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
} break;
case GGML_OP_COUNT_EQUAL:
case GGML_OP_SOLVE_TRI:
case GGML_OP_GATED_DELTA_NET:
{
n_tasks = n_threads;
} break;
@@ -2905,6 +2910,11 @@ struct ggml_cplan ggml_graph_plan(
{
cur = ggml_type_size(node->type)*(n_tasks + node->src[0]->ne[0]*n_tasks);
} break;
case GGML_OP_GATED_DELTA_NET:
{
const int64_t S_v = node->src[2]->ne[0];
cur = S_v * sizeof(float) * n_tasks;
} break;
case GGML_OP_COUNT:
{
GGML_ABORT("fatal error");