CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-17 11:37:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : remove implicit recurrent state rollbacks

Browse Source
This commit is contained in:
Francis Couture-Harpin
2024-11-24 20:35:30 -05:00
parent 124c222f76
commit 8006f3b3c8
25 changed files with 411 additions and 1119 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
examples/gritlm/gritlm.cpp
View File

@@ -44,7 +44,7 @@ static std::vector<std::vector<float>> encode(llama_context * ctx, const std::ve
}
// clear previous kv_cache values (irrelevant for embeddings)
llama_past_clear(ctx);
llama_kv_cache_clear(ctx);
llama_set_embeddings(ctx, true);
llama_set_causal_attn(ctx, false);
@@ -99,7 +99,7 @@ static std::string generate(llama_context * ctx, llama_sampler * smpl, const std
const llama_model * model = llama_get_model(ctx);
llama_token eos_token = llama_token_eos(model);
llama_past_clear(ctx);
llama_kv_cache_clear(ctx);
llama_set_embeddings(ctx, false);
llama_set_causal_attn(ctx, true);