Skip to content

Commit c0ba107

Browse files
committed
Update on "[ET-VK] Add reshape functions for transformers related operators"
## Changes * Implement resize functions for several operators used in Transformers models ## Motivation Be able to support batched prefill for llama models. Differential Revision: [D75686049](https://our.internmc.facebook.com/intern/diff/D75686049/) [ghstack-poisoned]
2 parents e196ed9 + cefb9df commit c0ba107

File tree

2 files changed

+1
-1
lines changed

2 files changed

+1
-1
lines changed

backends/vulkan/runtime/VulkanBackend.cpp

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -569,7 +569,6 @@ class VulkanBackend final : public ::executorch::runtime::BackendInterface {
569569

570570
if (should_propagate_resize) {
571571
compute_graph->propagate_resize();
572-
compute_graph->encode_execute();
573572
}
574573
compute_graph->execute();
575574

backends/vulkan/runtime/graph/ComputeGraph.cpp

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -712,6 +712,7 @@ void ComputeGraph::propagate_resize() {
712712
for (std::unique_ptr<ExecuteNode>& node : execute_nodes_) {
713713
node->trigger_resize(this);
714714
}
715+
encode_execute();
715716
}
716717

717718
} // namespace vkcompute

0 commit comments

Comments
 (0)