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

memory : rename interface to llama_memory_context_i (#14296)

Browse source 
* memory : rename interface to llama_memory_context_i

ggml-ci

* cont : fix comments

* cont : use "mctx" for referencing a memory context

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-06-21 08:03:46 +03:00 • committed by GitHub
parent b23fa0b3f4
commit 692e3cdd0a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
14 changed files with 339 additions and 341 deletions
Download patch file Download diff file Expand all files Collapse all files

53
src/llama-graph.h
View file

@ -17,12 +17,12 @@ struct ggml_tensor;
struct llama_ubatch;
struct llama_cparams;
struct llama_memory_state_i;
struct llama_memory_context_i;
class llama_kv_cache_unified_state;
class llama_kv_cache_unified_iswa_state;
class llama_memory_recurrent_state;
class llama_memory_hybrid_state;
class llama_kv_cache_unified_context;
class llama_kv_cache_unified_iswa_context;
class llama_memory_recurrent_context;
class llama_memory_hybrid_context;
// certain models (typically multi-modal) can produce different types of graphs
enum llm_graph_type {
@ -136,7 +136,7 @@ class llm_graph_input_pos_bucket_kv : public llm_graph_input_i {
public:
llm_graph_input_pos_bucket_kv(
const llama_hparams & hparams,
const llama_kv_cache_unified_state * kv_state) : hparams(hparams), kv_state(kv_state) {}
const llama_kv_cache_unified_context * mctx) : hparams(hparams), mctx(mctx) {}
virtual ~llm_graph_input_pos_bucket_kv() = default;
void set_input(const llama_ubatch * ubatch) override;
@ -144,7 +144,8 @@ public:
ggml_tensor * pos_bucket = nullptr; // I32 [n_kv, n_batch]
const llama_hparams & hparams;
const llama_kv_cache_unified_state * kv_state;
const llama_kv_cache_unified_context * mctx;
};
class llm_graph_input_out_ids : public llm_graph_input_i {
@ -191,14 +192,14 @@ public:
class llm_graph_input_rs : public llm_graph_input_i {
public:
llm_graph_input_rs(const llama_memory_recurrent_state * mem_state) : mem_state(mem_state) {}
llm_graph_input_rs(const llama_memory_recurrent_context * mctx) : mctx(mctx) {}
virtual ~llm_graph_input_rs() = default;
void set_input(const llama_ubatch * ubatch) override;
ggml_tensor * s_copy; // I32 [kv_size]
const llama_memory_recurrent_state * mem_state;
const llama_memory_recurrent_context * mctx;
};
class llm_graph_input_cross_embd : public llm_graph_input_i {
@ -238,10 +239,10 @@ public:
llm_graph_input_attn_kv_unified(
const llama_hparams & hparams,
const llama_cparams & cparams,
const llama_kv_cache_unified_state * kv_state) :
const llama_kv_cache_unified_context * mctx) :
hparams(hparams),
cparams(cparams),
kv_state(kv_state) {
mctx(mctx) {
}
~llm_graph_input_attn_kv_unified() = default;
@ -255,7 +256,7 @@ public:
const llama_hparams & hparams;
const llama_cparams & cparams;
const llama_kv_cache_unified_state * kv_state;
const llama_kv_cache_unified_context * mctx;
};
class llm_graph_input_attn_kv_unified_iswa : public llm_graph_input_i {
@ -263,10 +264,10 @@ public:
llm_graph_input_attn_kv_unified_iswa(
const llama_hparams & hparams,
const llama_cparams & cparams,
const llama_kv_cache_unified_iswa_state * kv_state) :
const llama_kv_cache_unified_iswa_context * mctx) :
hparams(hparams),
cparams(cparams),
kv_state(kv_state) {
mctx(mctx) {
}
~llm_graph_input_attn_kv_unified_iswa() = default;
@ -283,7 +284,7 @@ public:
const llama_hparams & hparams;
const llama_cparams & cparams;
const llama_kv_cache_unified_iswa_state * kv_state;
const llama_kv_cache_unified_iswa_context * mctx;
};
class llm_graph_input_attn_cross : public llm_graph_input_i {
@ -306,10 +307,10 @@ public:
llm_graph_input_mem_hybrid(
const llama_hparams & hparams,
const llama_cparams & cparams,
const llama_memory_hybrid_state * mem_state) :
const llama_memory_hybrid_context * mctx) :
hparams(hparams),
cparams(cparams),
mem_state(mem_state) {
mctx(mctx) {
}
virtual ~llm_graph_input_mem_hybrid() = default;
@ -325,7 +326,7 @@ public:
const llama_hparams & hparams;
const llama_cparams & cparams;
const llama_memory_hybrid_state * mem_state;
const llama_memory_hybrid_context * mctx;
};
//
@ -401,10 +402,10 @@ struct llm_graph_params {
ggml_backend_sched_t sched;
ggml_backend_t backend_cpu;
const llama_adapter_cvec * cvec;
const llama_adapter_loras * loras;
const llama_memory_state_i * mstate;
const llama_cross * cross;
const llama_adapter_cvec * cvec;
const llama_adapter_loras * loras;
const llama_memory_context_i * mctx;
const llama_cross * cross;
uint32_t n_outputs;
@ -453,10 +454,10 @@ struct llm_graph_context {
ggml_backend_t backend_cpu; // TODO: needed by build_attn_mha, figure out a way to remove?
const llama_adapter_cvec * cvec;
const llama_adapter_loras * loras;
const llama_memory_state_i * mstate;
const llama_cross * cross;
const llama_adapter_cvec * cvec;
const llama_adapter_loras * loras;
const llama_memory_context_i * mctx;
const llama_cross * cross;
const llm_graph_cb & cb_func;