tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-07-13 14:29:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

recurrent : call balloc split_reset() in init_batch() (#14414)

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2025-06-27 17:55:45 +03:00
committed by GitHub
parent 8d94219a4a
commit 43678060c1
1 changed files with 21 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/llama-memory-recurrent.cpp
View File

@ -363,8 +363,10 @@ llama_pos llama_memory_recurrent::seq_pos_max(llama_seq_id seq_id) const {
}
llama_memory_context_ptr llama_memory_recurrent::init_batch(llama_batch_allocr & balloc, uint32_t n_ubatch, bool embd_all) {
std::vector<llama_ubatch> ubatches;
do {
balloc.split_reset();
std::vector<llama_ubatch> ubatches;
while (true) {
llama_ubatch ubatch;
@ -383,10 +385,13 @@ llama_memory_context_ptr llama_memory_recurrent::init_batch(llama_batch_allocr &
}
if (!prepare(ubatches)) {
return std::make_unique<llama_memory_recurrent_context>(LLAMA_MEMORY_STATUS_FAILED_PREPARE);
break;
}
return std::make_unique<llama_memory_recurrent_context>(this, std::move(ubatches));
} while (false);
return std::make_unique<llama_memory_recurrent_context>(LLAMA_MEMORY_STATUS_FAILED_PREPARE);
}
llama_memory_context_ptr llama_memory_recurrent::init_full() {