vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-10 12:11:08 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge commit 'c342c3b93d' into concedo_experimental

Browse source 
# Conflicts:
#	.github/workflows/build.yml
#	CODEOWNERS
#	scripts/sync_vendor.py
This commit is contained in:
Concedo 2026-02-05 22:23:05 +08:00
commit 157fac7bd0
6 changed files with 330 additions and 77 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/models/qwen3next.cpp
View file

@ -265,13 +265,13 @@ std::pair<ggml_tensor *, ggml_tensor *> llm_build_qwen3next::build_delta_net_chu
cb(g_diff, "g_diff", il); // shape: (chunk_size, 1, n_chunks, H_v * n_seqs)
ggml_tensor * g_diff_exp = ggml_exp(ctx0, g_diff);
ggml_tensor * g_diff_exp_t = ggml_reshape_4d(ctx0, g_diff_exp,
ggml_tensor * g_diff_exp_t = ggml_reshape_4d(ctx0, g_diff_exp,
1, chunk_size, n_chunks, g_diff_exp->ne[3]);
ggml_tensor * key_gdiff = ggml_mul(ctx0, k, g_diff_exp_t);
cb(key_gdiff, "key_gdiff", il); // shape: (S_k, chunk_size, n_chunks, H_v * n_seqs)
ggml_tensor * key_gdiff_t = ggml_cont(ctx0, ggml_transpose(ctx0, key_gdiff));
ggml_tensor * key_gdiff_t = ggml_cont(ctx0, ggml_transpose(ctx0, key_gdiff));
cb(key_gdiff_t, "key_gdiff_t", il); // shape: (chunk_size, S_k, n_chunks, H_v * n_seqs)