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

llama : fix handling of "future" tokens when loading sessions

Browse source
This commit is contained in:
Georgi Gerganov 2023-10-03 18:29:22 +03:00
parent 0f332a9104
commit 337120cc0d
No known key found for this signature in database
GPG key ID: 449E073F9DC10735
6 changed files with 41 additions and 40 deletions
Download patch file Download diff file Expand all files Collapse all files

2
examples/server/server.cpp
View file

@ -448,7 +448,7 @@ struct llama_server_context
n_past = common_part(embd, prompt_tokens);
// since #3228 we now have to manually manage the KV cache
llama_kv_cache_seq_rm(ctx, 0, n_past, params.n_ctx);
llama_kv_cache_seq_rm(ctx, 0, n_past, -1);
embd = prompt_tokens;
if (n_past == num_prompt_tokens)