happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

readded cacheless logic

This commit is contained in:
ryan-mangeno 2025-10-29 12:29:44 -04:00
parent 04167d9e26
commit 7e8e1a07d5
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/llama-model.cpp
View File

@ -8221,7 +8221,7 @@ struct llm_build_modern_bert : public llm_graph_context {
ggml_tensor * inp_out_ids = build_inp_out_ids();
auto * inp_attn = build_attn_inp_kv_iswa(); // TODO: support cacheless iSWA embeddings [TAG_NO_CACHE_ISWA]
auto * inp_attn = build_attn_inp_no_cache();
for (int il = 0; il < n_layer; ++il) {
ggml_tensor * cur = inpL;
@ -19831,7 +19831,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
case LLM_ARCH_NOMIC_BERT_MOE:
case LLM_ARCH_NEO_BERT:
case LLM_ARCH_WAVTOKENIZER_DEC:
//case LLM_ARCH_MODERN_BERT: // TODO: disabled until cacheless SWA logic is fixed [TAG_NO_CACHE_ISWA]
case LLM_ARCH_MODERN_BERT:
case LLM_ARCH_GEMMA_EMBEDDING:
case LLM_ARCH_DREAM:
case LLM_ARCH_LLADA: