Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 17:25:07 +02:00
Code Issues Packages Projects Releases Wiki Activity

Revert "cleanup round 2"

Browse Source 
This reverts commit c7e39f9c97.
This commit is contained in:
Concedo
2026-06-09 22:33:00 +08:00
parent da0ba5cefd
commit ac8e77d82e
1 changed files with 4 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gpttype_adapter.cpp
+4 -11
View File
@@ -4512,9 +4512,6 @@ static void PrepareMediaEmbds(const int nctx, const std::vector<int> & media_int
{
printf("\nMTMD Media %i used Tokens: %d, Positions: %d, Boundary Tokens: %d",i,mediatokensneeded,mediaposneeded,boundarytokensneeded);
}
// Media chunks have two sizes: placeholder/embed tokens and KV positions.
// Text wrapper tokens consume both equally, so add them after taking the
// larger media-side requirement.
int mediactxneeded = std::max(mediatokensneeded, mediaposneeded) + boundarytokensneeded;
if(i==0)
{
@@ -5028,15 +5025,11 @@ generation_outputs gpttype_generate(const generation_inputs inputs)
}
//shorten memory if needed
const int media_context_size = std::max((int)last_media_mem.size(), last_media_pos_count);
const int media_context_extra = media_context_size - (int)last_media_mem.size();
if (embd_inp_mem.size() + media_context_extra + kcpp_data->n_predict + 4 > nctx)
if (embd_inp_mem.size() + kcpp_data->n_predict + 4 > nctx)
{
int limit = nctx - (kcpp_data->n_predict + 4) - media_context_extra;
const int protected_media_prefix = (int)last_media_mem.size() + (bos.empty() ? 0 : 1);
limit = std::max(limit, protected_media_prefix);
if (embd_inp_mem.size() > (size_t)limit) {
embd_inp_mem.resize((size_t)limit);
int limit = nctx - (kcpp_data->n_predict + 4);
if (embd_inp_mem.size() > limit) {
embd_inp_mem.resize(limit);
}
}
}