CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-13 10:57:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

models : move build_inp_out_ids outside loop (#17151)

Browse Source 
* move build_inp_out_ids outside loop

* realign
This commit is contained in:
Sigbjørn Skjæret
2025-11-10 22:55:30 +01:00
committed by GitHub
parent 395e286bc9
commit 7bef684118
2 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/models/ernie4-5.cpp
View File

@@ -1,7 +1,5 @@
#include "models.h" #include "models.h"
llm_build_ernie4_5::llm_build_ernie4_5(const llama_model & model, const llm_graph_params & params) : llm_build_ernie4_5::llm_build_ernie4_5(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params) { llm_graph_context(params) {
const int64_t n_embd_head = hparams.n_embd_head_v; const int64_t n_embd_head = hparams.n_embd_head_v;
@@ -19,6 +17,8 @@ llm_build_ernie4_5::llm_build_ernie4_5(const llama_model & model, const llm_grap
auto * inp_attn = build_attn_inp_kv(); auto * inp_attn = build_attn_inp_kv();
ggml_tensor * inp_out_ids = build_inp_out_ids();
for (int il = 0; il < n_layer; ++il) { for (int il = 0; il < n_layer; ++il) {
ggml_tensor * inpSA = inpL; ggml_tensor * inpSA = inpL;
@@ -67,7 +67,6 @@ llm_build_ernie4_5::llm_build_ernie4_5(const llama_model & model, const llm_grap
} }
if (il == n_layer - 1) { if (il == n_layer - 1) {
// skip computing output for unused tokens // skip computing output for unused tokens
ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids); cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
} }

3
src/models/openai-moe-iswa.cpp
View File

@@ -11,6 +11,8 @@ llm_build_openai_moe_iswa::llm_build_openai_moe_iswa(const llama_model & model,
auto * inp_attn = build_attn_inp_kv_iswa(); auto * inp_attn = build_attn_inp_kv_iswa();
ggml_tensor * inp_out_ids = build_inp_out_ids();
for (int il = 0; il < n_layer; ++il) { for (int il = 0; il < n_layer; ++il) {
ggml_tensor * inpSA = inpL; ggml_tensor * inpSA = inpL;
@@ -69,7 +71,6 @@ llm_build_openai_moe_iswa::llm_build_openai_moe_iswa(const llama_model & model,
} }
if (il == n_layer - 1) { if (il == n_layer - 1) {
// skip computing output for unused tokens // skip computing output for unused tokens
ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids); cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
} }