nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 4 Packages Projects Releases Wiki Activity

return filter to save memory (#24125)

Browse Source 
Co-authored-by: lvyichen <lvyichen@stepfun.com>
This commit is contained in:
forforever73
2026-06-04 21:56:33 +08:00
committed by GitHub
parent e8023568d0
commit 0dbfa66a1f
1 changed files with 9 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/llama-model.cpp
View File

@@ -2112,6 +2112,15 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
filter = [n_main](int32_t il) { return (uint32_t)il >= n_main; };
}
if (arch == LLM_ARCH_STEP35 && hparams.nextn_predict_layers > 0) {
const uint32_t n_main = hparams.n_layer - hparams.nextn_predict_layers;
if (params.ctx_type == LLAMA_CONTEXT_TYPE_MTP) {
filter = [n_main](int32_t il) { return (uint32_t)il >= n_main; };
} else {
filter = [n_main](int32_t il) { return (uint32_t)il < n_main; };
}
}
if (hparams.swa_type != LLAMA_SWA_TYPE_NONE) {
GGML_ASSERT(hparams.is_swa_any());