CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-04 09:32:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

refactor: Remove n_embd_k/v_s from unified cache

Browse Source 
No longer needed now that unified isn't also supporting recurrent

https://github.com/ggml-org/llama.cpp/pull/13979#discussion_r2140761069

Branch: HybridRecurrentCache
This commit is contained in:
Gabe Goodhart
2025-06-11 12:20:04 -06:00
parent 833dfb54ae
commit 1dd12133cd
1 changed files with 8 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
src/llama-kv-cache-unified.cpp
View File

@@ -68,8 +68,8 @@ llama_kv_cache_unified::llama_kv_cache_unified(
continue; continue;
} }
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(il); const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il);
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(il); const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il);
const char * dev_name = "CPU"; const char * dev_name = "CPU";
@@ -1430,7 +1430,7 @@ void llama_kv_cache_unified::state_write_data(llama_io_write_i & io, const std::
for (const auto & layer : layers) { for (const auto & layer : layers) {
const uint32_t il = layer.il; const uint32_t il = layer.il;
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(il); const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il);
// Write key type // Write key type
const int32_t k_type_i = (int32_t)layer.k->type; const int32_t k_type_i = (int32_t)layer.k->type;
@@ -1452,7 +1452,7 @@ void llama_kv_cache_unified::state_write_data(llama_io_write_i & io, const std::
for (const auto & layer : layers) { for (const auto & layer : layers) {
const uint32_t il = layer.il; const uint32_t il = layer.il;
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(il); const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il);
// Write value type // Write value type
const int32_t v_type_i = (int32_t)layer.v->type; const int32_t v_type_i = (int32_t)layer.v->type;
@@ -1476,7 +1476,7 @@ void llama_kv_cache_unified::state_write_data(llama_io_write_i & io, const std::
for (const auto & layer : layers) { for (const auto & layer : layers) {
const uint32_t il = layer.il; const uint32_t il = layer.il;
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(il); const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il);
// Write value type // Write value type
const int32_t v_type_i = (int32_t)layer.v->type; const int32_t v_type_i = (int32_t)layer.v->type;
@@ -1621,7 +1621,7 @@ bool llama_kv_cache_unified::state_read_data(llama_io_read_i & io, uint32_t cell
for (const auto & layer : layers) { for (const auto & layer : layers) {
const uint32_t il = layer.il; const uint32_t il = layer.il;
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(il); const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il);
// Read type of key // Read type of key
int32_t k_type_i_ref; int32_t k_type_i_ref;
@@ -1651,7 +1651,7 @@ bool llama_kv_cache_unified::state_read_data(llama_io_read_i & io, uint32_t cell
for (const auto & layer : layers) { for (const auto & layer : layers) {
const uint32_t il = layer.il; const uint32_t il = layer.il;
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(il); const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il);
// Read type of value // Read type of value
int32_t v_type_i_ref; int32_t v_type_i_ref;
@@ -1681,7 +1681,7 @@ bool llama_kv_cache_unified::state_read_data(llama_io_read_i & io, uint32_t cell
for (const auto & layer : layers) { for (const auto & layer : layers) {
const uint32_t il = layer.il; const uint32_t il = layer.il;
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(il); const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il);
// Read type of value // Read type of value
int32_t v_type_i_ref; int32_t v_type_i_ref;