swa full used unless ctx shift and fast forward disabled

This commit is contained in:
Concedo 2025-05-21 22:47:45 +08:00
commit 9f976e9c65
16 changed files with 1429 additions and 654 deletions

View file

@ -1144,6 +1144,7 @@ struct llama_context_params common_context_params_to_llama(const common_params &
cparams.flash_attn = params.flash_attn;
cparams.no_perf = params.no_perf;
cparams.op_offload = !params.no_op_offload;
cparams.swa_full = params.swa_full;
if (params.reranking) {
cparams.embeddings = true;