vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 00:54:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into concedo_experimental

Browse source 
# Conflicts:
#	Makefile
#	README.md
#	flake.lock
#	ggml-cuda.cu
#	llama.cpp
#	tests/test-backend-ops.cpp
#	tests/test-quantize-fns.cpp
This commit is contained in:
Concedo 2024-02-28 13:41:35 +08:00
commit ad638285de
26 changed files with 3393 additions and 589 deletions
Download patch file Download diff file Expand all files Collapse all files

2
gpttype_adapter.cpp
View file

@ -666,7 +666,7 @@ void PurgeMissingTokens(llama_context * ctx, std::vector<int> &current_context_t
//extract the unwanted tokens out from context and KV
int diff = found - trimstart;
llama_kv_cache_seq_rm(llama_ctx_v4, 0, trimstart, trimstart + diff);
llama_kv_cache_seq_shift(llama_ctx_v4, 0, trimstart + diff, -1, -diff);
llama_kv_cache_seq_add(llama_ctx_v4, 0, trimstart + diff, -1, -diff);
for (size_t i = trimstart + diff; i < current_context_tokens.size() - 1; i++)
{