sdgoij/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-03-17 16:44:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

wip

Browse Source
This commit is contained in:
Georgi Gerganov
2026-02-10 23:44:42 +02:00
parent 0cc02542a8
commit 5372fc6461
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/models/qwen3next.cpp
View File

@@ -225,8 +225,8 @@ std::pair<ggml_tensor *, ggml_tensor *> llm_build_qwen3next::build_delta_net_chu
cb(k_cumdecay, "k_cumdecay", il); // shape: (chunk_size, chunk_size, n_chunks, H_v * n_seqs)
ggml_tensor * attn_kq = ggml_mul_mat(ctx0, k, q);
attn_kq = ggml_mul(ctx0, decay_mask, attn_kq);
attn_kq = ggml_mul(ctx0, attn_kq, diag_mask);
attn_kq = ggml_mul(ctx0, attn_kq, decay_mask);
attn_kq = ggml_mul(ctx0, attn_kq, diag_mask);
cb(attn_kq, "attn_kq", il); // shape: (chunk_size, chunk_size, n_chunks, H_v * n_seqs)
// vectorized calculation of key_gdiff