Test DNN created from apptainer image

cms-sw · Dec 19, 2023 · 07c40c0 · 07c40c0
1 parent 4579ebc
commit 07c40c0
Show file tree

Hide file tree

Showing 7 changed files with 97 additions and 35 deletions.
diff --git a/PhysicsTools/PyTorch/test/BuildFile.xml b/PhysicsTools/PyTorch/test/BuildFile.xml
@@ -22,3 +22,39 @@
     <use name="HeterogeneousCore/CUDAUtilities"/>
   </bin>
 </iftool>
+
+
+<iftool name="cuda">
+  <bin name="testTorchFromBufferPinnedMemory" file="testTorchFromBufferPinnedMemory.cu">
+    <use name="boost_filesystem"/>
+    <use name="catch2"/>
+    <use name="cppunit"/>
+    <use name="cuda"/>
+    <use name="pytorch"/>
+    <use name="FWCore/ParameterSet"/>
+    <use name="FWCore/ParameterSetReader"/>
+    <use name="FWCore/PluginManager"/>
+    <use name="FWCore/ServiceRegistry"/>
+    <use name="FWCore/Utilities"/>
+    <use name="HeterogeneousCore/CUDAServices"/>
+    <use name="HeterogeneousCore/CUDAUtilities"/>
+  </bin>
+</iftool>
+
+
+<iftool name="cuda">
+  <bin name="testTorchFromBufferModelEval" file="testTorchFromBufferModelEval.cu">
+    <use name="boost_filesystem"/>
+    <use name="catch2"/>
+    <use name="cppunit"/>
+    <use name="cuda"/>
+    <use name="pytorch"/>
+    <use name="FWCore/ParameterSet"/>
+    <use name="FWCore/ParameterSetReader"/>
+    <use name="FWCore/PluginManager"/>
+    <use name="FWCore/ServiceRegistry"/>
+    <use name="FWCore/Utilities"/>
+    <use name="HeterogeneousCore/CUDAServices"/>
+    <use name="HeterogeneousCore/CUDAUtilities"/>
+  </bin>
+</iftool>
diff --git a/PhysicsTools/PyTorch/test/create_simple_dnn.py b/PhysicsTools/PyTorch/test/create_simple_dnn.py
@@ -1,5 +1,16 @@
+import sys
+import os
 import torch
 
+# prepare the datadir
+if len(sys.argv) >= 2:
+    datadir = sys.argv[1]
+else:
+    thisdir = os.path.dirname(os.path.abspath(__file__))
+    datadir = os.path.join(os.path.dirname(thisdir), "bin", "data")
+
+os.makedirs(datadir, exist_ok=True)
+
 class MyModule(torch.nn.Module):
     def __init__(self, N, M):
         super(MyModule, self).__init__()
@@ -15,6 +26,6 @@ def forward(self, input):
 
 tm = torch.jit.trace(module.eval(), x)
 
-print(tm.graph)
+tm.save(f"{datadir}/simple_dnn.pt")
 
-tm.save("simple_dnn.pt")
+print("simple_dnn.pt created successfully!")
diff --git a/PhysicsTools/PyTorch/test/simple_dnn.pt b/PhysicsTools/PyTorch/test/simple_dnn.pt
diff --git a/PhysicsTools/PyTorch/test/testBase.h b/PhysicsTools/PyTorch/test/testBase.h
@@ -14,21 +14,38 @@
 class testBasePyTorch : public CppUnit::TestFixture {
 public:
   std::string dataPath_;
-  std::string testPath_;
 
   void setUp();
   void tearDown();
   std::string cmsswPath(std::string path);
 
   virtual void test() = 0;
+
+  virtual std::string pyScript() const = 0;
 };
 
 void testBasePyTorch::setUp() {
   dataPath_ =
       cmsswPath("/test/" + std::string(std::getenv("SCRAM_ARCH")) + "/" + boost::filesystem::unique_path().string());
 
-  // create the graph
-  testPath_ = cmsswPath("/src/PhysicsTools/PyTorch/test");
+  // create the graph using apptainer
+  std::string testPath = cmsswPath("/src/PhysicsTools/PyTorch/test");
+  std::string cmd = "apptainer exec -B " + cmsswPath("") +
+                    "  /cvmfs/unpacked.cern.ch/registry.hub.docker.com/cmsml/cmsml:3.11  python " + testPath + "/" +
+                    pyScript() + " " + dataPath_;
+  std::cout << "cmd: " << cmd << std::endl;
+  std::array<char, 128> buffer;
+  std::string result;
+  std::shared_ptr<FILE> pipe(popen(cmd.c_str(), "r"), pclose);
+  if (!pipe) {
+    throw std::runtime_error("popen() failed!");
+  }
+  while (!feof(pipe.get())) {
+    if (fgets(buffer.data(), 128, pipe.get()) != NULL) {
+      result += buffer.data();
+    }
+  }
+  std::cout << std::endl << result << std::endl;
 }
 
 void testBasePyTorch::tearDown() {

diff --git a/PhysicsTools/PyTorch/test/testBaseCUDA.h b/PhysicsTools/PyTorch/test/testBaseCUDA.h
@@ -27,23 +27,39 @@
 class testBasePyTorchCUDA : public CppUnit::TestFixture {
 public:
   std::string dataPath_;
-  std::string testPath_;
 
   void setUp();
   void tearDown();
   std::string cmsswPath(std::string path);
 
+  virtual std::string pyScript() const = 0;
+
   virtual void test() = 0;
 };
 
 void testBasePyTorchCUDA::setUp() {
   dataPath_ =
       cmsswPath("/test/" + std::string(std::getenv("SCRAM_ARCH")) + "/" + boost::filesystem::unique_path().string());
 
-  // create the graph
-  testPath_ = cmsswPath("/src/PhysicsTools/PyTorch/test");
+  // create the graph using apptainer
+  std::string testPath = cmsswPath("/src/PhysicsTools/PyTorch/test");
+  std::string cmd = "apptainer exec -B " + cmsswPath("") +
+                    "  /cvmfs/unpacked.cern.ch/registry.hub.docker.com/cmsml/cmsml:3.11  python " + testPath + "/" +
+                    pyScript() + " " + dataPath_;
+  std::cout << "cmd: " << cmd << std::endl;
+  std::array<char, 128> buffer;
+  std::string result;
+  std::shared_ptr<FILE> pipe(popen(cmd.c_str(), "r"), pclose);
+  if (!pipe) {
+    throw std::runtime_error("popen() failed!");
+  }
+  while (!feof(pipe.get())) {
+    if (fgets(buffer.data(), 128, pipe.get()) != NULL) {
+      result += buffer.data();
+    }
+  }
+  std::cout << std::endl << result << std::endl;
 }
-
 void testBasePyTorchCUDA::tearDown() {
   if (std::filesystem::exists(dataPath_)) {
     std::filesystem::remove_all(dataPath_);

diff --git a/PhysicsTools/PyTorch/test/testTorchSimpleDnn.cc b/PhysicsTools/PyTorch/test/testTorchSimpleDnn.cc
@@ -10,13 +10,16 @@ class testSimpleDNN : public testBasePyTorch {
   CPPUNIT_TEST_SUITE_END();
 
 public:
+  std::string pyScript() const override;
   void test() override;
 };
 
 CPPUNIT_TEST_SUITE_REGISTRATION(testSimpleDNN);
 
+std::string testSimpleDNN::pyScript() const { return "create_simple_dnn.py"; }
+
 void testSimpleDNN::test() {
-  std::string model_path = testPath_ + "/simple_dnn.pt";
+  std::string model_path = dataPath_ + "/simple_dnn.pt";
   torch::Device device(torch::kCPU);
   torch::jit::script::Module module;
   try {
@@ -36,15 +39,3 @@ void testSimpleDNN::test() {
   CPPUNIT_ASSERT(output.item<float_t>() == 110.);
   std::cout << "ok\n";
 }
-
-// int main(int argc, const char* argv[]) {
-//   std::cout << "Running model on CPU" << std::endl;
-//   torch::Device cpu(torch::kCPU);
-//   runModel("/data/user/dvalsecc/simple_dnn.pt", cpu);
-
-//   std::cout << "Running model on CUDA" << std::endl;
-//   torch::Device cuda(torch::kCUDA);
-//   runModel("/data/user/dvalsecc/simple_dnn.pt", cuda);
-
-//   return 0;
-// }
diff --git a/PhysicsTools/PyTorch/test/testTorchSimpleDnnCUDA.cc b/PhysicsTools/PyTorch/test/testTorchSimpleDnnCUDA.cc
@@ -10,11 +10,14 @@ class testSimpleDNNCUDA : public testBasePyTorchCUDA {
   CPPUNIT_TEST_SUITE_END();
 
 public:
+  std::string pyScript() const override;
   void test() override;
 };
 
 CPPUNIT_TEST_SUITE_REGISTRATION(testSimpleDNNCUDA);
 
+std::string testSimpleDNNCUDA::pyScript() const { return "create_simple_dnn.py"; }
+
 void testSimpleDNNCUDA::test() {
   if (!cms::cudatest::testDevices())
     return;
@@ -40,7 +43,7 @@ process.add_(cms.Service('CUDAService'))
 
   std::cout << "Testing CUDA backend" << std::endl;
 
-  std::string model_path = testPath_ + "/simple_dnn.pt";
+  std::string model_path = dataPath_ + "/simple_dnn.pt";
   torch::Device device(torch::kCUDA);
   torch::jit::script::Module module;
   try {
@@ -60,15 +63,3 @@ process.add_(cms.Service('CUDAService'))
   CPPUNIT_ASSERT(output.item<float_t>() == 110.);
   std::cout << "ok\n";
 }
-
-// int main(int argc, const char* argv[]) {
-//   std::cout << "Running model on CPU" << std::endl;
-//   torch::Device cpu(torch::kCPU);
-//   runModel("/data/user/dvalsecc/simple_dnn.pt", cpu);
-
-//   std::cout << "Running model on CUDA" << std::endl;
-//   torch::Device cuda(torch::kCUDA);
-//   runModel("/data/user/dvalsecc/simple_dnn.pt", cuda);
-
-//   return 0;
-// }