tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-27 03:55:20 +00:00
Code Issues Packages Projects Releases 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
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");

17
tools/server/server.cpp
View File

@ -1137,9 +1137,6 @@ struct server_task_result_metrics : server_task_result {
int n_tasks_deferred;
int64_t t_start;
int32_t kv_cache_tokens_count;
int32_t kv_cache_used_cells;
// TODO: somehow reuse server_metrics in the future, instead of duplicating the fields
uint64_t n_prompt_tokens_processed_total = 0;
uint64_t t_prompt_processing_total = 0;
@ -1179,9 +1176,6 @@ struct server_task_result_metrics : server_task_result {
{ "n_decode_total", n_decode_total },
{ "n_busy_slots_total", n_busy_slots_total },
{ "kv_cache_tokens_count", kv_cache_tokens_count },
{ "kv_cache_used_cells", kv_cache_used_cells },
{ "slots", slots_data },
};
}
@ -2771,9 +2765,6 @@ struct server_context {
res->n_tasks_deferred = queue_tasks.queue_tasks_deferred.size();
res->t_start = metrics.t_start;
res->kv_cache_tokens_count = llama_kv_self_n_tokens(ctx);
res->kv_cache_used_cells = llama_kv_self_used_cells(ctx);
res->n_prompt_tokens_processed_total = metrics.n_prompt_tokens_processed_total;
res->t_prompt_processing_total = metrics.t_prompt_processing_total;
res->n_tokens_predicted_total = metrics.n_tokens_predicted_total;
@ -3883,14 +3874,6 @@ int main(int argc, char ** argv) {
{"name", "predicted_tokens_seconds"},
{"help", "Average generation throughput in tokens/s."},
{"value", res_metrics->n_tokens_predicted ? 1.e3 / res_metrics->t_tokens_generation * res_metrics->n_tokens_predicted : 0.}
},{
{"name", "kv_cache_usage_ratio"},
{"help", "KV-cache usage. 1 means 100 percent usage."},
{"value", 1. * res_metrics->kv_cache_used_cells / params.n_ctx}
},{
{"name", "kv_cache_tokens"},
{"help", "KV-cache tokens."},
{"value", (uint64_t) res_metrics->kv_cache_tokens_count}
},{
{"name", "requests_processing"},
{"help", "Number of requests processing."},