tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-27 03:55:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

context : allow cache-less context for embeddings (#13108)

Browse Source 
* context : allow cache-less context for embeddings

ggml-ci

* context : enable reranking with encode()

ggml-ci

* context : encode() clears embd_seq

ggml-ci

* examples : use llama_encode() when appropriate

ggml-ci

* models : nomic bert moe does not require KV cache

* llama : update comments for llama_decode/llama_encode

ggml-ci

* context : update warning log [no ci]
This commit is contained in:
Georgi Gerganov
2025-05-08 14:28:33 +03:00
committed by GitHub
parent 51fb96b1ff
commit 6562e5a4d6
5 changed files with 47 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
examples/embedding/embedding.cpp
View File

@ -35,23 +35,14 @@ static void batch_add_seq(llama_batch & batch, const std::vector<int32_t> & toke
static void batch_decode(llama_context * ctx, llama_batch & batch, float * output, int n_seq, int n_embd, int embd_norm) {
const enum llama_pooling_type pooling_type = llama_pooling_type(ctx);
const struct llama_model * model = llama_get_model(ctx);
// clear previous kv_cache values (irrelevant for embeddings)
llama_kv_self_clear(ctx);
// run model
LOG_INF("%s: n_tokens = %d, n_seq = %d\n", __func__, batch.n_tokens, n_seq);
if (llama_model_has_encoder(model) && !llama_model_has_decoder(model)) {
// encoder-only model
if (llama_encode(ctx, batch) < 0) {
LOG_ERR("%s : failed to encode\n", __func__);
}
} else if (!llama_model_has_encoder(model) && llama_model_has_decoder(model)) {
// decoder-only model
if (llama_decode(ctx, batch) < 0) {
LOG_ERR("%s : failed to decode\n", __func__);
}
if (llama_encode(ctx, batch) < 0) {
LOG_ERR("%s : failed to encode\n", __func__);
}
for (int i = 0; i < batch.n_tokens; i++) {