ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

llama : rework embeddings logic (#14208)

Browse source 
* llama : rework embeddings logic

ggml-ci

* cont : fix rerank

ggml-ci

* cont : engrish [no ci]

* cont : fix rerank

ggml-ci

* server : support both embeddings and completions with single model

ggml-ci

* cont : avoid embeddings_org

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-06-16 14:14:00 +03:00 • committed by GitHub
parent 3ba0d843c6
commit d3e64b9f49
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
16 changed files with 159 additions and 114 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/llama-kv-cache-unified-iswa.cpp
View file

@ -95,8 +95,8 @@ llama_pos llama_kv_cache_unified_iswa::seq_pos_max(llama_seq_id seq_id) const {
return kv_swa->seq_pos_max(seq_id);
}
llama_memory_state_ptr llama_kv_cache_unified_iswa::init_batch(const llama_batch & batch, uint32_t n_ubatch, bool embd_pooled) {
GGML_UNUSED(embd_pooled);
llama_memory_state_ptr llama_kv_cache_unified_iswa::init_batch(const llama_batch & batch, uint32_t n_ubatch, bool embd_all) {
GGML_UNUSED(embd_all);
// first try simple split
do {