Skip to content

Commit bf31654

Browse files
committed
kompute : disable LLAMA_SPLIT_LAYER after ggml-org#5321
Signed-off-by: Jared Van Bortel <[email protected]>
1 parent fa654d0 commit bf31654

File tree

1 file changed

+4
-1
lines changed

1 file changed

+4
-1
lines changed

llama.cpp

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -3647,6 +3647,7 @@ static bool llm_load_tensors(
36473647
model.buft_layer[i] = llama_default_buffer_type_cpu(true);
36483648
}
36493649

3650+
#ifndef GGML_USE_KOMPUTE
36503651
if (split_mode == LLAMA_SPLIT_LAYER) {
36513652
// calculate the split points
36523653
int device_count = llama_get_device_count();
@@ -3684,7 +3685,9 @@ static bool llm_load_tensors(
36843685
} else {
36853686
model.buft_output = llama_default_buffer_type_cpu(true);
36863687
}
3687-
} else {
3688+
} else
3689+
#endif
3690+
{
36883691
ggml_backend_buffer_type_t split_buft;
36893692
if (split_mode == LLAMA_SPLIT_ROW) {
36903693
split_buft = llama_default_buffer_type_split(main_gpu, tensor_split);

0 commit comments

Comments
 (0)