ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-01-26 17:20:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

OK, this seems to be working

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-12-16 18:40:53 +00:00
parent 2fcf407ab3
commit 7a7029ec99
2 changed files with 2 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/ggml-backend.cpp
View File

@@ -2012,11 +2012,7 @@ static void ggml_backend_sched_copy_inputs(ggml_backend_sched_t sched, ggml_back
// try async copy, but if not possible, we can still use a sync copy without synchronizing the dst backend, since we handle the synchronization here with multiple copies and events
// TODO: add public function to facilitate this, since applications do not have direct access to the backend interface
if (!split_backend->iface.cpy_tensor_async || !split_backend->iface.cpy_tensor_async(input_backend, split_backend, input, input_cpy)) {
int input_backend_id = tensor_backend_id(input);
if (needs_sync[input_backend_id]) {
ggml_backend_synchronize(input_backend);
needs_sync[input_backend_id] = k_set_sync;
}
ggml_backend_synchronize(input_backend);
if (needs_sync[split_backend_id]) {
if (sched->events[split_backend_id][sched->cur_copy] != NULL) {
ggml_backend_event_synchronize(sched->events[split_backend_id][sched->cur_copy]);

2
src/llama.cpp
View File

@@ -4773,7 +4773,7 @@ struct llama_context * llama_new_context_with_model(
LLAMA_LOG_INFO("XXXXXXXXXXXXXXXXXXXXX Setting only active experts offload\n");
ggml_backend_sched_set_only_active_experts(ctx->sched, true);
}
if (model->split_mode == LLAMA_SPLIT_MODE_GRAPH && !model->has_tensor_overrides()) {
if (model->split_mode == LLAMA_SPLIT_MODE_GRAPH) { // && !model->has_tensor_overrides()) {
ggml_backend_sched_set_split_mode_graph(ctx->sched, true);
ggml_backend_sched_set_max_extra_alloc(ctx->sched, params.max_extra_alloc);
}