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

Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843)

Browse source 
* Extend llama_kv_cache_seq_rm to allow matichng any sequence

* Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear

Use llama_kv_cache_clear for cache clearing

Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
This commit is contained in:
Kerfuffle 2023-10-29 11:31:40 -06:00 • committed by GitHub
parent 2046eb4345
commit 6e08281e58
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
8 changed files with 30 additions and 32 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/common.cpp
View file

@ -889,7 +889,7 @@ std::tuple<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
std::vector<llama_token> tmp = { llama_token_bos(model), llama_token_eos(model), };
llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch), 0, 0));
llama_kv_cache_tokens_rm(lctx, -1, -1);
llama_kv_cache_clear(lctx);
llama_reset_timings(lctx);
}