tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-16 13:12:51 -04:00
Code Issues Packages Projects Releases Wiki Activity

llama : add high-throughput mode (#14363)

Browse Source 
* kv-cache : prepare K/V buffers for separation

ggml-ci

* batched-bench : fix oob write

ggml-ci

* llama : add "virtual sequences"

ggml-ci

* llama : use "stream" vs "virtual sequence"

ggml-ci

* graph : fix stream splitting when KV cache is not used

ggml-ci

* kv-cache : add multi-stream save/load support

ggml-ci

* llama : add "--attn-streams" flag

ggml-ci

* kv-cache : fix handling when find_slot fails

ggml-ci

* kv-cache : restore find_slot impl

ggml-ci

* kv-cache : add comments

* kv-cache : add bounds checks for sequence id

ggml-ci

* cont : add n_seq_max to batch allocr

ggml-ci

* kv-cache : perform stream copies lazily after llama_synchronize

ggml-ci

* kv-cache : avoid throwing exceptions across the C boundary

ggml-ci

* CUDA: 4D FlashAttention support (#14628)

* CUDA: 4D FlashAttention support

* CUDA: fix WMMA FA kernel

* llama : rename attn_streams -> kv_unified

ggml-ci

* common : rename kv_split -> kv_unified

ggml-ci

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
Georgi Gerganov
2025-07-16 16:35:42 +03:00
committed by GitHub
parent ab14019821
commit 225e7a1438
30 changed files with 1080 additions and 460 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -18,16 +18,17 @@ llama_kv_cache_unified_iswa::llama_kv_cache_unified_iswa(
bool v_trans,
bool offload,
bool swa_full,
bool unified,
uint32_t kv_size,
uint32_t n_seq_max,
uint32_t n_ubatch,
uint32_t n_pad) : hparams(model.hparams) {
uint32_t n_pad) : hparams(model.hparams), unified(unified) {
llama_kv_cache_unified::layer_filter_cb filter_base = [&](int32_t il) { return !model.hparams.is_swa(il); };
llama_kv_cache_unified::layer_filter_cb filter_swa = [&](int32_t il) { return model.hparams.is_swa(il); };
const uint32_t size_base = kv_size;
uint32_t size_swa = std::min(size_base, GGML_PAD(hparams.n_swa*n_seq_max + n_ubatch, n_pad));
uint32_t size_swa = std::min(size_base, GGML_PAD(hparams.n_swa*(unified ? n_seq_max : 1) + n_ubatch, n_pad));
// when using full-size SWA cache, we set the SWA cache size to be equal to the base cache size
if (swa_full) {
@@ -41,14 +42,14 @@ llama_kv_cache_unified_iswa::llama_kv_cache_unified_iswa(
kv_base = std::make_unique<llama_kv_cache_unified>(
model, std::move(filter_base), type_k, type_v,
v_trans, offload, size_base, n_seq_max, n_pad,
v_trans, offload, unified, size_base, n_seq_max, n_pad,
0, LLAMA_SWA_TYPE_NONE);
LLAMA_LOG_INFO("%s: creating SWA KV cache, size = %u cells\n", __func__, size_swa);
kv_swa = std::make_unique<llama_kv_cache_unified>(
model, std::move(filter_swa), type_k, type_v,
v_trans, offload, size_swa, n_seq_max, n_pad,
v_trans, offload, unified, size_swa, n_seq_max, n_pad,
hparams.n_swa, hparams.swa_type);
}
@@ -100,6 +101,11 @@ llama_memory_context_ptr llama_kv_cache_unified_iswa::init_batch(llama_batch_all
// first try simple split
do {
if (!unified) {
// requires equal splits, so we skip the simple split
break;
}
balloc.split_reset();
std::vector<llama_ubatch> ubatches;
@@ -140,7 +146,7 @@ llama_memory_context_ptr llama_kv_cache_unified_iswa::init_batch(llama_batch_all
std::vector<llama_ubatch> ubatches;
while (true) {
auto ubatch = balloc.split_equal(n_ubatch, false);
auto ubatch = balloc.split_equal(n_ubatch, !unified);
if (ubatch.n_tokens == 0) {
break;