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

use vector for query_emb

Browse source
This commit is contained in:
Minsoo Cheong 2024-03-24 21:13:38 +09:00
parent 5d02fe94bd
commit 3073b380cb
1 changed files with 3 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

7
examples/retrieval/retrieval.cpp
View file

@ -314,10 +314,9 @@ int main(int argc, char ** argv) {
struct llama_batch query_batch = llama_batch_init(n_batch, 0, 1); struct llama_batch query_batch = llama_batch_init(n_batch, 0, 1);
batch_add_seq(query_batch, query_tokens, 0); batch_add_seq(query_batch, query_tokens, 0);
float * query_emb = new float[n_embd]; std::vector<float> query_emb(n_embd, 0);
batch_decode(ctx, query_batch, query_emb, 1, n_embd); batch_decode(ctx, query_batch, query_emb.data(), 1, n_embd);
std::vector<float> query_embedding(query_emb, query_emb + n_embd); std::vector<float> query_embedding(query_emb.data(), query_emb.data() + n_embd);
delete[] query_emb;
llama_batch_clear(query_batch); llama_batch_clear(query_batch);
// similarities tuple vector (chunk index, similarity) // similarities tuple vector (chunk index, similarity)