starred/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-04-25 02:06:11 +02:00
Code Issues Packages Projects Releases Wiki Activity

Revert "Update vendored llama.cpp to b7847" (#14061)

Browse Source
This commit is contained in:
Jeffrey Morgan
2026-02-03 18:39:36 -08:00
committed by GitHub
parent a6355329bf
commit b1fccabb34
240 changed files with 5050 additions and 21247 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
llama/llama.cpp/src/models/openai-moe-iswa.cpp vendored
View File

@@ -14,9 +14,6 @@ llm_build_openai_moe_iswa::llm_build_openai_moe_iswa(const llama_model & model,
ggml_tensor * inp_out_ids = build_inp_out_ids();
for (int il = 0; il < n_layer; ++il) {
const float freq_base_l = model.get_rope_freq_base (cparams, il);
const float freq_scale_l = model.get_rope_freq_scale(cparams, il);
ggml_tensor * inpSA = inpL;
// norm
@@ -52,13 +49,13 @@ llm_build_openai_moe_iswa::llm_build_openai_moe_iswa(const llama_model & model,
Qcur = ggml_rope_ext(
ctx0, Qcur, inp_pos, nullptr,
n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
Kcur = ggml_rope_ext(
ctx0, Kcur, inp_pos, nullptr,
n_rot, rope_type, n_ctx_orig, freq_base_l, freq_scale_l,
n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);