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

kv-cache : separate recurrent vs non-recurrent impl (#12799)

Browse source 
* kv-cache : serparate recurrent vs non-recurrent impl (wip)

ggml-ci

* kv-cache : init -> contructor + add llama_memory_params

ggml-ci

* kv-cache : fix callback reference

ggml-ci

* context : llama_kv_cache -> llama_memory_i

ggml-ci

* context : move memory creation logic to model

ggml-ci

* llama : remove reference of memory during encode

ggml-ci

* kv-cache : hide padding details in the implementation

ggml-ci

* kv-cache : add ubatch_next()

ggml-ci

* context : simplify sbatch logic

ggml-ci

* kv-cache : hide defrag logic in the implementation

ggml-ci

* context : hide kv cache details in implementation

ggml-ci

* build : fix

ggml-ci

* cont : another fix

ggml-ci

* kv-cache : simplify interface (wip)

ggml-ci

* kv-cache : use separate KV cell structs for unified/recurrent

ggml-ci

* kv-cache : clean-up

ggml-ci

* model : better llama_model::create_model() signature

ggml-ci

* kv-cache : fix recurrent seq_rm()

ggml-ci

* kv-cache : replace `struct callbacks` with `llama_model &`

ggml-ci

* kv-cache : replace `struct graph_params` with `llama_context &`

ggml-ci

* kv-cache : fix offload check

ggml-ci

* context : avoid passing unique_ptr

ggml-ci

* kv-cache : avoid using the backends from the llama_context

ref #13113

ggml-ci

* kv-cache : more consistent debug logs [no ci]

* kv-cache : do not pass the full llama_context for kv graphs

ggml-ci

* kv-cache : remove comment

* kv-cache : ggml_rope_ext_inplace -> ggml_rope_ext

ggml-ci

* kv-cache : fix recurrent multi-user case

ggml-ci

* memory : remove comments [no ci]
This commit is contained in:
Georgi Gerganov 2025-05-02 17:48:36 +03:00 • committed by GitHub
parent cb06a3c363
commit c642bc014c
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
11 changed files with 1960 additions and 1048 deletions
Download patch file Download diff file Expand all files Collapse all files

45
src/llama-context.h
View file

@ -27,7 +27,12 @@ struct llama_context {
void synchronize();
const llama_model & get_model() const;
const llama_model & get_model() const;
const llama_cparams & get_cparams() const;
ggml_backend_sched_t get_sched() const;
ggml_context * get_ctx_compute() const;
uint32_t n_ctx() const;
uint32_t n_ctx_per_seq() const;
@ -137,49 +142,30 @@ private:
// Returns max number of outputs for which space was reserved.
int32_t output_reserve(int32_t n_outputs);
// make the outputs have the same order they had in the user-provided batch
// TODO: maybe remove this
void output_reorder();
//
// graph
//
public:
int32_t graph_max_nodes() const;
// zero-out inputs and create the ctx_compute for the compute graph
ggml_cgraph * graph_init();
llm_graph_result_ptr graph_build(
ggml_context * ctx,
ggml_cgraph * gf,
const llama_ubatch & ubatch,
llm_graph_type gtype);
// returns the result of ggml_backend_sched_graph_compute_async execution
ggml_status graph_compute(
ggml_cgraph * gf,
bool batched);
private:
llm_graph_result_ptr graph_build(
ggml_context * ctx,
ggml_cgraph * gf,
const llama_ubatch & ubatch,
llm_graph_type gtype);
llm_graph_cb graph_get_cb() const;
// used by kv_self_update()
ggml_tensor * build_rope_shift(
ggml_context * ctx0,
ggml_tensor * cur,
ggml_tensor * shift,
ggml_tensor * factors,
float freq_base,
float freq_scale) const;
llm_graph_result_ptr build_kv_self_shift(
ggml_context * ctx0,
ggml_cgraph * gf) const;
llm_graph_result_ptr build_kv_self_defrag(
ggml_context * ctx0,
ggml_cgraph * gf) const;
// TODO: read/write lora adapters and cvec
size_t state_write_data(llama_io_write_i & io);
size_t state_read_data (llama_io_read_i & io);
@ -196,11 +182,10 @@ private:
llama_cparams cparams;
llama_adapter_cvec cvec;
llama_adapter_loras loras;
llama_sbatch sbatch;
llama_cross cross; // TODO: tmp for handling cross-attention - need something better probably
std::unique_ptr<llama_kv_cache_unified> kv_self;
std::unique_ptr<llama_memory_i> memory;
// TODO: remove
bool logits_all = false;