ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

kv-cache : simplify the interface (#13660)

Browse source 
* kv-cache : simplify the interface

ggml-ci

* context : revert llama_batch_allocr position change

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-05-21 15:11:13 +03:00 • committed by GitHub
parent b44890df2e
commit 797f2ac062
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
9 changed files with 89 additions and 153 deletions
Download patch file Download diff file Expand all files Collapse all files

4
tools/run/run.cpp
View file

@ -936,7 +936,7 @@ static int apply_chat_template(const struct common_chat_templates * tmpls, Llama
// Function to tokenize the prompt
static int tokenize_prompt(const llama_vocab * vocab, const std::string & prompt,
std::vector<llama_token> & prompt_tokens, const LlamaData & llama_data) {
const bool is_first = llama_kv_self_used_cells(llama_data.context.get()) == 0;
const bool is_first = llama_kv_self_seq_pos_max(llama_data.context.get(), 0) == 0;
const int n_prompt_tokens = -llama_tokenize(vocab, prompt.c_str(), prompt.size(), NULL, 0, is_first, true);
prompt_tokens.resize(n_prompt_tokens);
@ -952,7 +952,7 @@ static int tokenize_prompt(const llama_vocab * vocab, const std::string & prompt
// Check if we have enough space in the context to evaluate this batch
static int check_context_size(const llama_context_ptr & ctx, const llama_batch & batch) {
const int n_ctx = llama_n_ctx(ctx.get());
const int n_ctx_used = llama_kv_self_used_cells(ctx.get());
const int n_ctx_used = llama_kv_self_seq_pos_max(ctx.get(), 0);
if (n_ctx_used + batch.n_tokens > n_ctx) {
printf(LOG_COL_DEFAULT "\n");
printe("context size exceeded\n");