Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-19 09:15:04 +02:00
Code Issues Packages Projects Releases Wiki Activity

spec: add eagle3-v3 support for gpt-oss model (#25794)

Browse Source
This commit is contained in:
Ruixiang Wang
2026-07-28 08:58:16 +02:00
committed by GitHub
parent c6292cfb8e
commit f87067841b
9 changed files with 59 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/models/openai-moe.cpp
+7 -1
View File
@@ -116,7 +116,7 @@ llama_model_openai_moe::graph::graph(const llama_model & model, const llm_graph_
cb(cur, "attn_out", il);
}
if (il == n_layer - 1) {
if (il == n_layer - 1 && inp_out_ids && cparams.embeddings_nextn_masked) {
// skip computing output for unused tokens
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
@@ -154,6 +154,12 @@ llama_model_openai_moe::graph::graph(const llama_model & model, const llm_graph_
}
cur = inpL;
res->t_h_nextn = cur;
if (!cparams.embeddings_nextn_masked && inp_out_ids) {
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
}
cur = build_norm(cur,
model.output_norm, NULL,
LLM_NORM_RMS, -1);