Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-18 00:35:04 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/upstream' into concedo_experimental

Browse Source 
# Conflicts:
#	.github/workflows/build.yml
#	.github/workflows/release.yml
#	examples/eval-callback/eval-callback.cpp
#	ggml/src/ggml-cann/common.h
#	ggml/src/ggml-vulkan/CMakeLists.txt
#	ggml/src/ggml-vulkan/vulkan-shaders/CMakeLists.txt
#	tests/test-backend-ops.cpp
This commit is contained in:
Concedo
2025-06-28 17:47:53 +08:00
parent 4ec0e0fd21 b25e92774e
commit a88c56e70c
15 changed files with 593 additions and 227 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-memory-recurrent.cpp
+21 -16
View File
@@ -363,30 +363,35 @@ 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();
while (true) {
llama_ubatch ubatch;
std::vector<llama_ubatch> ubatches;
while (true) {
llama_ubatch ubatch;
if (embd_all) {
// if all tokens are output, split by sequence
ubatch = balloc.split_seq(n_ubatch);
} else {
ubatch = balloc.split_equal(n_ubatch);
if (embd_all) {
// if all tokens are output, split by sequence
ubatch = balloc.split_seq(n_ubatch);
} else {
ubatch = balloc.split_equal(n_ubatch);
}
if (ubatch.n_tokens == 0) {
break;
}
ubatches.push_back(std::move(ubatch)); // NOLINT
}
if (ubatch.n_tokens == 0) {
if (!prepare(ubatches)) {
break;
}
ubatches.push_back(std::move(ubatch)); // NOLINT
}
return std::make_unique<llama_memory_recurrent_context>(this, std::move(ubatches));
} while (false);
if (!prepare(ubatches)) {
return std::make_unique<llama_memory_recurrent_context>(LLAMA_MEMORY_STATUS_FAILED_PREPARE);
}
return std::make_unique<llama_memory_recurrent_context>(this, std::move(ubatches));
return std::make_unique<llama_memory_recurrent_context>(LLAMA_MEMORY_STATUS_FAILED_PREPARE);
}
llama_memory_context_ptr llama_memory_recurrent::init_full() {