saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-29 20:14:01 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Set correct ep_cache_context

Browse source
This commit is contained in:
Jingyan Wang 2025-01-15 00:10:37 +00:00
parent a63973e2e7
commit 0ac2ac716b
1 changed files with 6 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

16
onnxruntime/core/providers/tensorrt/tensorrt_execution_provider.cc
View file

@ -3358,10 +3358,9 @@ Status TensorrtExecutionProvider::CreateNodeComputeInfoFromGraph(const GraphView
// dump EP context node model
if (dump_ep_context_model_) {
// "ep_cache_context" node attribute should be a relative path to context model directory
if (ep_cache_context_attr_.empty()) {
auto cache_file_name = std::filesystem::path(engine_cache_path).filename();
ep_cache_context_attr_ = std::filesystem::path(engine_cache_relative_path_to_context_model_dir).append(cache_file_name.string()).string();
}
// ep_cache_context_attr_ needs to be set to engine_cache_path for every context node
auto cache_file_name = std::filesystem::path(engine_cache_path).filename();
ep_cache_context_attr_ = std::filesystem::path(engine_cache_relative_path_to_context_model_dir).append(cache_file_name.string()).string();
std::string compute_capability_hw_compat = compute_capability_;
if (engine_cache_enable_ && engine_hw_compatible_) {
compute_capability_hw_compat = "80+";
@ -3375,7 +3374,6 @@ Status TensorrtExecutionProvider::CreateNodeComputeInfoFromGraph(const GraphView
compute_capability_hw_compat,
model_path_,
GetLogger());
auto& graph = trt_ep_context_model_ptr->MainGraph();
trt_ep_context_models.emplace_back(std::move(trt_ep_context_model_ptr));
}
}
@ -3466,10 +3464,9 @@ Status TensorrtExecutionProvider::CreateNodeComputeInfoFromGraph(const GraphView
// However, if the embed_mode is 0 (only includes engine path), TRT EP will serialize it here.
if (dump_ep_context_model_ && has_dynamic_shape) {
// "ep_cache_context" node attribute should be a relative path to context model directory
if (ep_cache_context_attr_.empty()) {
auto cache_file_name = std::filesystem::path(engine_cache_path).filename();
ep_cache_context_attr_ = std::filesystem::path(engine_cache_relative_path_to_context_model_dir).append(cache_file_name.string()).string();
}
// ep_cache_context_attr_ needs to be set to engine_cache_path for every context node
auto cache_file_name = std::filesystem::path(engine_cache_path).filename();
ep_cache_context_attr_ = std::filesystem::path(engine_cache_relative_path_to_context_model_dir).append(cache_file_name.string()).string();
std::string compute_capability_hw_compat = compute_capability_;
if (engine_cache_enable_ && engine_hw_compatible_) {
compute_capability_hw_compat = "80+";
@ -4410,7 +4407,6 @@ const InlinedVector<const Node*> TensorrtExecutionProvider::GetEpContextNodes()
for (const auto& context_model: trt_ep_context_models) {
const auto& graph = context_model->MainGraph();
for (const auto& node: graph.Nodes()) {
// if (node.IsEpContextNode()) { // Check if it's an EP context node
ep_context_nodes.push_back(node);
}
}