Skip to content

[ML] Add pytorch_process rss memory stat to inference response #2896

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

Open
darius-vil wants to merge 2 commits into
base: main
Choose a base branch
from
Open

[ML] Add pytorch_process rss memory stat to inference response #2896

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
9646a68
Add memory stats to pytorch inference response
darius-vil Jan 27, 2026
9d8df85
Add changelog
darius-vil Jan 27, 2026
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
8 changes: 7 additions & 1 deletion bin/pytorch_inference/CResultWriter.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,8 @@ void CResultWriter::writeError(const std::string_view& requestId, const std::str
void CResultWriter::wrapAndWriteInnerResponse(const std::string& innerResponse,
const std::string& requestId,
bool isCacheHit,
std::uint64_t timeMs) {
std::uint64_t timeMs,
std::size_t residentSize) {
core::CBoostJsonConcurrentLineWriter jsonWriter{m_WrappedOutputStream};
jsonWriter.onObjectBegin();
jsonWriter.onKey(CCommandParser::REQUEST_ID);
Expand All @@ -67,6 +68,11 @@ void CResultWriter::wrapAndWriteInnerResponse(const std::string& innerResponse,
jsonWriter.onBool(isCacheHit);
jsonWriter.onKey(TIME_MS);
jsonWriter.onUint64(timeMs);
jsonWriter.onKey(PROCESS_STATS);
jsonWriter.onObjectBegin();
jsonWriter.onKey(MEMORY_RESIDENT_SET_SIZE);
jsonWriter.onUint64(residentSize);
jsonWriter.onObjectEnd();
jsonWriter.rawKeyAndValue(innerResponse);
jsonWriter.onObjectEnd();
}
Expand Down
3 changes: 2 additions & 1 deletion bin/pytorch_inference/CResultWriter.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,7 +83,8 @@ class CResultWriter : public TStringBufWriter {
void wrapAndWriteInnerResponse(const std::string& innerResponse,
const std::string& requestId,
bool isCacheHit,
std::uint64_t timeMs);
std::uint64_t timeMs,
std::size_t residentSize);

//! Write the prediction portion of an inference result.
template<std::size_t N>
Expand Down
8 changes: 5 additions & 3 deletions bin/pytorch_inference/Main.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -133,9 +133,11 @@ bool handleRequest(ml::torch::CCommandParser::CRequestCacheInterface& cache,
}
},
[&](const auto& innerResponseJson_, bool isCacheHit) {
resultWriter.wrapAndWriteInnerResponse(innerResponseJson_,
requestId, isCacheHit,
stopWatch.stop());
std::size_t residentSetSize =
ml::core::CProcessStats::residentSetSize();
resultWriter.wrapAndWriteInnerResponse(
innerResponseJson_, requestId, isCacheHit,
stopWatch.stop(), residentSetSize);
});
});
return true;
Expand Down
9 changes: 5 additions & 4 deletions bin/pytorch_inference/unittest/CResultWriterTest.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,11 +99,12 @@ BOOST_AUTO_TEST_CASE(testWrapAndWriteInferenceResult) {
std::ostringstream output;
{
ml::torch::CResultWriter resultWriter{output};
resultWriter.wrapAndWriteInnerResponse(innerPortion, "req4", true, 123);
resultWriter.wrapAndWriteInnerResponse(innerPortion, "req4", true, 123, 111);
}
std::string expected = "[{\"request_id\":\"req4\",\"cache_hit\":true,"
"\"time_ms\":123,\"result\":{\"inference\":"
"[[[1.0,1.0,1.0],[1.0,1.0,1.0],[1.0,1.0,1.0],[1.0,1.0,1.0],[1.0,1.0,1.0]]]}}\n]";
std::string expected =
"[{\"request_id\":\"req4\",\"cache_hit\":true,"
"\"time_ms\":123,\"process_stats\":{\"memory_rss\":111},\"result\":{\"inference\":"
"[[[1.0,1.0,1.0],[1.0,1.0,1.0],[1.0,1.0,1.0],[1.0,1.0,1.0],[1.0,1.0,1.0]]]}}\n]";
std::string actual = output.str();

LOG_INFO(<< "expected: " << expected);
Expand Down
1 change: 1 addition & 0 deletions docs/CHANGELOG.asciidoc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,7 @@
=== Enhancements

* Downgrade log severity for a batch of recoverable errors. (See {ml-pull}[#2889].)
* Add pytorch_process rss memory stat to inference response. (See {ml-pull}[#2896].)

== {es} version 9.2.0

Expand Down