tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-04 08:15:55 -04:00
Code Issues Packages Projects Releases Wiki Activity

kv-cache : use ggml_set_rows (#14285)

Browse Source 
* kv-cache : use ggml_set_rows

ggml-ci

* graph : separate k and v indices

ggml-ci

* cont : remove redundant ifs

ggml-ci

* kv-cache : improve find_slot impl

* kv-cache : bounds-check when accessing slot_info indices

* kv-cache : add comments

ggml-ci

* ggml : add TODOs for adding GGML_OP_SET_ROWS support in the backends

ggml-ci
This commit is contained in:
Georgi Gerganov
2025-07-03 10:53:35 +03:00
committed by GitHub
parent 9067487c44
commit a70c8a0c4b
13 changed files with 451 additions and 143 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
src/llama-graph.h
View File

@@ -249,8 +249,14 @@ public:
void set_input(const llama_ubatch * ubatch) override;
ggml_tensor * get_k_idxs() const { return self_k_idxs; }
ggml_tensor * get_v_idxs() const { return self_v_idxs; }
ggml_tensor * get_kq_mask() const { return self_kq_mask_cnv; }
ggml_tensor * self_k_idxs = nullptr; // I64 [n_batch]
ggml_tensor * self_v_idxs = nullptr; // I64 [n_batch]
ggml_tensor * self_kq_mask = nullptr; // F32 [n_kv, n_batch]
ggml_tensor * self_kq_mask_cnv = nullptr; // [n_kv, n_batch]
@@ -274,9 +280,19 @@ public:
void set_input(const llama_ubatch * ubatch) override;
ggml_tensor * get_k_idxs() const { return self_k_idxs; }
ggml_tensor * get_v_idxs() const { return self_v_idxs; }
ggml_tensor * get_k_idxs_swa() const { return self_k_idxs_swa; }
ggml_tensor * get_v_idxs_swa() const { return self_v_idxs_swa; }
ggml_tensor * get_kq_mask() const { return self_kq_mask_cnv; }
ggml_tensor * get_kq_mask_swa() const { return self_kq_mask_swa_cnv; }
ggml_tensor * self_k_idxs = nullptr; // I64 [n_batch]
ggml_tensor * self_v_idxs = nullptr; // I64 [n_batch]
ggml_tensor * self_k_idxs_swa = nullptr; // I64 [n_batch]
ggml_tensor * self_v_idxs_swa = nullptr; // I64 [n_batch]
ggml_tensor * self_kq_mask = nullptr; // F32 [n_kv, n_batch]
ggml_tensor * self_kq_mask_cnv = nullptr; // [n_kv, n_batch]
ggml_tensor * self_kq_mask_swa = nullptr; // F32 [n_kv, n_batch]
@@ -319,8 +335,14 @@ public:
ggml_tensor * s_copy; // I32 [kv_size]
ggml_tensor * get_k_idxs() const { return self_k_idxs; }
ggml_tensor * get_v_idxs() const { return self_v_idxs; }
ggml_tensor * get_kq_mask() const { return self_kq_mask_cnv; }
ggml_tensor * self_k_idxs = nullptr; // I64 [n_batch]
ggml_tensor * self_v_idxs = nullptr; // I64 [n_batch]
ggml_tensor * self_kq_mask = nullptr; // F32 [n_kv, n_batch]
ggml_tensor * self_kq_mask_cnv = nullptr; // [n_kv, n_batch]
@@ -336,7 +358,7 @@ public:
llm_graph_input_one() {}
virtual ~llm_graph_input_one() = default;
void set_input(const llama_ubatch *) override;
void set_input(const llama_ubatch * ubatch) override;
ggml_tensor * one = nullptr; // F32
};