vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

batch : add n_used count (#14512)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2025-07-04 09:04:59 +03:00 • committed by GitHub
parent 499a8f5a78
commit c79184d2d1
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 34 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

9
src/llama-batch.cpp
View file

@ -405,6 +405,10 @@ uint32_t llama_batch_allocr::get_n_outputs() const {
return n_outputs;
}
uint32_t llama_batch_allocr::get_n_used() const {
return n_used;
}
std::vector<int32_t> & llama_batch_allocr::get_out_ids() {
return out_ids;
}
@ -420,6 +424,8 @@ llama_pos llama_batch_allocr::seq_pos_max(llama_seq_id seq_id) const {
void llama_batch_allocr::split_reset() {
out_ids.clear();
n_used = 0;
used.clear();
used.resize(get_n_tokens(), false);
@ -444,6 +450,7 @@ llama_ubatch llama_batch_allocr::split_simple(uint32_t n_ubatch) {
idxs.push_back(cur_idx);
used[cur_idx] = true;
++n_used;
++cur_idx;
@ -529,6 +536,7 @@ llama_ubatch llama_batch_allocr::split_equal(uint32_t n_ubatch) {
idxs_per_seq[s].push_back(idx);
used[idx] = true;
++n_used;
++cur_idx[s];
}
@ -570,6 +578,7 @@ llama_ubatch llama_batch_allocr::split_seq(uint32_t n_ubatch) {
idxs.push_back(cur_idx);
used[cur_idx] = true;
++n_used;
if (idxs.size() >= n_ubatch) {
break;