Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 1 addition & 2 deletions onnxruntime/core/providers/openvino/ov_interface.cc
Original file line number Diff line number Diff line change
Expand Up @@ -552,7 +552,6 @@ void StatefulOVInferRequest::RewindKVCache(size_t index) {
}

OVTensorPtr StatefulOVInferRequest::GetTensor(const std::string& input_name) {

auto tobj = OVInferRequest::GetTensor(input_name);

if (_npu_logits_slice_required) {
Expand Down Expand Up @@ -583,7 +582,7 @@ OVTensorPtr StatefulOVInferRequest::GetTensor(const std::string& input_name) {
return blob;
}
},
"Could not create sliced logits tensor");
"Could not create sliced logits tensor");
}
}
}
Expand Down
11 changes: 5 additions & 6 deletions onnxruntime/core/providers/openvino/ov_stateful_patch_utils.cc
Original file line number Diff line number Diff line change
Expand Up @@ -178,7 +178,6 @@ std::pair<std::vector<std::string>, std::unordered_set<std::string>> ExtractKVPa
// not_kv_inputs = ["input_ids", "attention_mask"]
std::pair<std::vector<std::string>, std::vector<std::string>> ExtractInputKVTensors(
const std::shared_ptr<ov::Model>& model, const std::unordered_set<std::string>& kv_patterns) {

std::vector<std::string> key_value_input_names;
std::vector<std::string> not_kv_inputs;

Expand Down Expand Up @@ -258,11 +257,11 @@ void PatchStatefulDecoder(std::shared_ptr<ov::Model> model) {
}

if (key_value_input_names.size() != key_value_output_names.size()) {
ORT_THROW("Found different sizes between key_value_input_names (",
key_value_input_names.size(),
") and key_value_output_names (",
key_value_output_names.size(),
"). They couldn't be paired.");
ORT_THROW("Found different sizes between key_value_input_names (",
key_value_input_names.size(),
") and key_value_output_names (",
key_value_output_names.size(),
"). They couldn't be paired.");
}

// By default, batch is the 0 - th but chatglm uses 1 - st dimension as batch
Expand Down
Loading