happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

graph : use fill instead of scale_bias in grouped expert selection (#17867) 

* use fill instead of scale_bias in grouped expert selection

* do not explicitly use _inplace
This commit is contained in:
Sigbjørn Skjæret 2025-12-08 21:29:59 +01:00 committed by GitHub
parent 2fa51c19b0
commit c8554b66e0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llama-graph.cpp
View File

@ -973,7 +973,7 @@ ggml_tensor * llm_graph_context::build_moe_ffn(
// mask out the other groups // mask out the other groups
selection_probs = ggml_get_rows(ctx0, selection_groups, expert_groups); // [n_exp_per_group, n_group_used, n_tokens] selection_probs = ggml_get_rows(ctx0, selection_groups, expert_groups); // [n_exp_per_group, n_group_used, n_tokens]
selection_probs = ggml_set_rows(ctx0, ggml_scale_bias(ctx0, selection_groups, 0.0f, -INFINITY), selection_probs, expert_groups); // [n_exp_per_group, n_expert_groups, n_tokens] selection_probs = ggml_set_rows(ctx0, ggml_fill(ctx0, selection_groups, -INFINITY), selection_probs, expert_groups); // [n_exp_per_group, n_expert_groups, n_tokens]
selection_probs = ggml_reshape_2d(ctx0, selection_probs, n_expert, n_tokens); // [n_expert, n_tokens] selection_probs = ggml_reshape_2d(ctx0, selection_probs, n_expert, n_tokens); // [n_expert, n_tokens]
cb(selection_probs, "ffn_moe_probs_masked", il); cb(selection_probs, "ffn_moe_probs_masked", il);
} }