Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

DeepseekV4: fix seq_rm (#25588)

Browse Source 
* DeepseekV4: fix seq_rm

* implement proper seq_cp

* create actual update context
This commit is contained in:
Aman Gupta
2026-07-14 21:45:36 +08:00
committed by GitHub
parent 7cbd61002d
commit 7f575c39d6
2 changed files with 106 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-kv-cache-dsv4.h
+21 -4
View File
@@ -10,6 +10,10 @@
class llama_dsv4_comp_state {
public:
using stream_copy_info = llama_kv_cache::stream_copy_info;
stream_copy_info sc_info;
llama_dsv4_comp_state(
const llama_model & model,
bool offload,
@@ -22,6 +26,8 @@ public:
const llama_memory_i::layer_filter_cb & filter);
void clear(llama_seq_id seq_id, bool data);
void seq_cp(llama_seq_id seq_id_src, llama_seq_id seq_id_dst);
void apply_copies(const stream_copy_info & sc_info) const;
uint32_t get_ratio() const;
uint32_t get_state_size() const;
@@ -44,6 +50,9 @@ private:
ggml_tensor * kv;
ggml_tensor * score;
std::vector<ggml_tensor *> kv_stream;
std::vector<ggml_tensor *> score_stream;
};
const uint32_t ratio;
@@ -245,6 +254,7 @@ private:
class llama_kv_cache_dsv4_context : public llama_memory_context_i {
public:
using slot_info_vec_t = llama_kv_cache::slot_info_vec_t;
using stream_copy_info = llama_kv_cache::stream_copy_info;
struct comp_plan {
// Per-ubatch recipe for updating compressor state, committing completed
@@ -291,7 +301,10 @@ public:
llama_kv_cache_dsv4_context(
llama_kv_cache_dsv4 * kv,
llama_context * lctx,
bool optimize);
bool optimize,
stream_copy_info sc_info_csa,
stream_copy_info sc_info_hca,
stream_copy_info sc_info_lid);
llama_kv_cache_dsv4_context(
llama_kv_cache_dsv4 * kv,
@@ -351,9 +364,13 @@ private:
const std::unique_ptr<llama_kv_cache_dsv4_comp_context> ctx_hca;
const std::unique_ptr<llama_kv_cache_dsv4_comp_context> ctx_lid;
const llama_dsv4_comp_state * csa_state = nullptr;
const llama_dsv4_comp_state * hca_state = nullptr;
const llama_dsv4_comp_state * lid_state = nullptr;
llama_dsv4_comp_state * csa_state = nullptr;
llama_dsv4_comp_state * hca_state = nullptr;
llama_dsv4_comp_state * lid_state = nullptr;
stream_copy_info sc_info_csa;
stream_copy_info sc_info_hca;
stream_copy_info sc_info_lid;
bool reserve_plans = false;
mutable comp_plan reserve_plan_csa;