tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-18 08:37:43 +00:00
Code Issues Packages Projects Releases Wiki Activity

context : prepare for abstraction

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2025-01-17 21:11:03 +02:00
parent 99422dfa3f
commit a0c500b4dc
3 changed files with 323 additions and 309 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
src/llama-context.h
View File

@ -30,11 +30,14 @@ struct llama_batch_manager_i {
virtual void finalize() = 0;
};
// TODO: make implementation details private
// TODO: become abstract base class, split the current implementation into different child classes
struct llama_context {
llama_context(const llama_model & model)
: model(model)
, t_start_us(model.t_start_us)
, t_load_us (model.t_load_us) {}
// TODO: store the worst-case graph build function and reuse it later
llama_context(
const llama_model & model,
const llama_context_params & params,
std::function<ggml_cgraph *(llama_context &, const llama_ubatch &)> fn_build_graph_worst);
const struct llama_model & model;