CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-08 10:07:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

context : fix n_ctx_per_seq computation

Browse Source
This commit is contained in:
Georgi Gerganov
2025-10-23 14:51:26 +03:00
parent a42fb77147
commit 492f628c58
2 changed files with 7 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llama-model.cpp
View File

@@ -6712,7 +6712,7 @@ float llama_model::get_rope_freq_scale(const llama_cparams & cparams, int il) co
}
ggml_tensor * llama_model::get_rope_factors(const llama_cparams & cparams, int il) const {
const uint32_t n_ctx_per_seq = cparams.n_ctx / cparams.n_seq_max;
const uint32_t n_ctx_per_seq = cparams.kv_unified ? cparams.n_ctx : cparams.n_ctx / cparams.n_seq_max;
// choose long/short freq factors based on the context size
if (layers[il].rope_freqs != nullptr) {