Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Move checkpoint to IterationData. Remove ExecIterData. #5555

Merged
merged 2 commits into from
Jul 10, 2024
Merged

Move checkpoint to IterationData. Remove ExecIterData. #5555

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
717a327
Move checkpoint to IterationData. Remove ExecIterData.
mzient Jul 9, 2024
523e527
API fix.
mzient Jul 9, 2024
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
17 changes: 9 additions & 8 deletions dali/pipeline/executor/executor_impl.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -157,7 +157,7 @@ void Executor<WorkspacePolicy, QueuePolicy>::SyncDevice() {

template <typename WorkspacePolicy, typename QueuePolicy>
void Executor<WorkspacePolicy, QueuePolicy>::RunCPUImpl(size_t iteration_id) {
GetCurrentIterationData(iteration_id).iter_data->iteration_index = iteration_id;
GetCurrentIterationData(iteration_id)->iteration_index = iteration_id;
PreRun();
const char placement_error[] =
"Cannot run a pipeline with Mixed/GPU ops in CPU-only mode. Please provide "
Expand Down Expand Up @@ -418,7 +418,7 @@ void Executor<WorkspacePolicy, QueuePolicy>::RunHelper(OpNode &op_node, Workspac
const auto &schema = spec.GetSchema();
SmallVector<int, 16> empty_layout_in_idxs;

ws.InjectIterationData(GetCurrentIterationData(iteration_id).iter_data);
ws.InjectIterationData(GetCurrentIterationData(iteration_id));
ws.ClearOperatorTraces();

auto ws_order = ws.has_stream() ? AccessOrder(ws.stream()) : AccessOrder::host();
Expand Down Expand Up @@ -599,7 +599,7 @@ void Executor<WorkspacePolicy, QueuePolicy>::ShareOutputsImpl(Workspace *ws, siz
), DALI_FAIL("Invalid storage device")); // NOLINT(whitespace/parens)
}

ws->InjectIterationData(GetCurrentIterationData(iteration_id).iter_data);
ws->InjectIterationData(GetCurrentIterationData(iteration_id));


// Mostly a sanity check - we don't want to return a non-contiguous batch to Python.
Expand Down Expand Up @@ -664,10 +664,11 @@ void Executor<WorkspacePolicy, QueuePolicy>::InitIterationData() {
size_t iteration_data_size = CalcIterationDataSize();
iteration_data_.resize(iteration_data_size);
for (auto& id : iteration_data_) {
id.iter_data = std::make_shared<IterationData>();
id = std::make_shared<IterationData>();
if (checkpointing_) {
id->checkpoint = std::make_shared<Checkpoint>();
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is it a problem that checkpoint would be NULL when checkpointing_ == false?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think so, @stiepan ?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Me neither.

for (auto &node : graph_->GetOpNodes())
id.checkpoint.AddOperator(node.instance_name);
id->checkpoint->AddOperator(node.instance_name);
}
}
}
Expand Down Expand Up @@ -707,7 +708,7 @@ template<typename WorkspacePolicy, typename QueuePolicy>
void Executor<WorkspacePolicy, QueuePolicy>::CreateCheckpoint(const OpNode &op_node,
int iteration_id,
AccessOrder order) {
auto &cpt = GetCurrentIterationData(iteration_id).checkpoint;
auto &cpt = *GetCurrentIterationData(iteration_id)->checkpoint;
auto &op_cpt = cpt.GetOpCheckpoint(op_node.id);
assert(op_cpt.OperatorName() == op_node.instance_name);
cpt.SetIterationId(iteration_id);
Expand Down Expand Up @@ -744,7 +745,7 @@ void Executor<WorkspacePolicy, QueuePolicy>::CreateInitialCheckpoints() {
template<typename WorkspacePolicy, typename QueuePolicy>
Checkpoint &Executor<WorkspacePolicy, QueuePolicy>::GetCurrentCheckpoint() {
DALI_ENFORCE(checkpointing_, "Cannot access checkpoints when checkpointing is not enabled. ");
auto &cpt = GetCurrentIterationData(output_iteration_id_).checkpoint;
auto &cpt = *GetCurrentIterationData(output_iteration_id_)->checkpoint;
return cpt;
}

Expand All @@ -767,7 +768,7 @@ void Executor<WorkspacePolicy, QueuePolicy>::RestoreStateFromCheckpoint(const Ch
}

template<typename WorkspacePolicy, typename QueuePolicy>
ExecIterData &
const SharedIterData &
Executor<WorkspacePolicy, QueuePolicy>::GetCurrentIterationData(size_t iteration_id) {
return iteration_data_[iteration_id % iteration_data_.size()];
}
Expand Down
6 changes: 3 additions & 3 deletions dali/pipeline/executor/executor_impl.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,6 @@
#include "dali/pipeline/executor/queue_metadata.h"
#include "dali/pipeline/executor/queue_policy.h"
#include "dali/pipeline/executor/workspace_policy.h"
#include "dali/pipeline/executor/iteration_data.h"
#include "dali/pipeline/executor/lowered_graph.h"
#include "dali/pipeline/operator/batch_size_provider.h"
#include "dali/pipeline/operator/builtin/conditional/split_merge.h"
Expand All @@ -48,6 +47,7 @@
#include "dali/pipeline/util/event_pool.h"
#include "dali/pipeline/util/stream_pool.h"
#include "dali/pipeline/util/thread_pool.h"
#include "dali/pipeline/workspace/iteration_data.h"
#include "dali/pipeline/workspace/workspace_data_factory.h"


Expand Down Expand Up @@ -339,7 +339,7 @@ class DLL_PUBLIC Executor : public ExecutorBase, public QueuePolicy {

WorkspacePolicy ws_policy_;

std::vector<ExecIterData> iteration_data_;
std::vector<SharedIterData> iteration_data_;
size_t cpu_iteration_id_ = 0, mixed_iteration_id_ = 0, gpu_iteration_id_ = 0;
size_t output_iteration_id_ = 0;

Expand Down Expand Up @@ -450,7 +450,7 @@ class DLL_PUBLIC Executor : public ExecutorBase, public QueuePolicy {
/**
* Returns the iteration data for given iteration ID and stage.
*/
virtual ExecIterData& GetCurrentIterationData(size_t iteration_id);
virtual const SharedIterData &GetCurrentIterationData(size_t iteration_id);
};

template <typename WorkspacePolicy, typename QueuePolicy>
Expand Down
33 changes: 0 additions & 33 deletions dali/pipeline/executor/iteration_data.h
View file
Open in desktop

This file was deleted.

5 changes: 5 additions & 0 deletions dali/pipeline/workspace/iteration_data.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,8 @@ namespace dali {
using operator_trace_map_t = std::unordered_map<
std::string /* trace_name */, std::string /* trace_value */>;

class Checkpoint;

class OperatorTraces {
public:
/** Gets operator traces for a single operator.
Expand Down Expand Up @@ -74,8 +76,11 @@ class OperatorTraces {
struct IterationData {
int64_t iteration_index = 0;
OperatorTraces operator_traces;
std::shared_ptr<Checkpoint> checkpoint;
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I used shared_ptr here because it allows Checkpoint to be only forward-declared. Using unique_ptr wouldn't be much more elegant and would require IterationData to be DLL_PUBLIC (meh) and a separate translation unit for the destructor (blah) or a unique_ptr with a custom deleter (blah).

};

using SharedIterData = std::shared_ptr<IterationData>;

} // namespace dali

#endif // DALI_PIPELINE_WORKSPACE_ITERATION_DATA_H_
Expand Down
8 changes: 4 additions & 4 deletions dali/pipeline/workspace/workspace.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -545,16 +545,16 @@ class WorkspaceBase : public ArgumentWorkspace {

///@{
/** Sets shared data associated with the current iteration */
void InjectIterationData(std::shared_ptr<IterationData> iter_data) {
void InjectIterationData(SharedIterData iter_data) {
if (iter_data != iter_data_) {
operator_traces_ = nullptr;
iter_data_ = std::move(iter_data);
}
}

/** Gets the shared data associated with the current iteration */
auto &GetIterationData() const {
return *iter_data_;
SharedIterData GetIterationData() const {
return iter_data_;
}

/**
Expand Down Expand Up @@ -705,7 +705,7 @@ class WorkspaceBase : public ArgumentWorkspace {
mutable operator_trace_map_t *operator_traces_ = nullptr;

/** Data shared across all workspaces in the current iteration. */
std::shared_ptr<IterationData> iter_data_;
SharedIterData iter_data_;
};

class Workspace : public WorkspaceBase<TensorList> {};
Expand Down
Loading