rpms/root
4
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
root/root-tmva-Speed-up-TMVA-CNN-and-RNN-tutorials.patch
Mattias Ellert a08fcbc5ad Update to 6.28.00 
ROOT now uses llvm/clang version 13 (updated from version 9)
Clean up specfile by removing EPEL 7 conditionals
Drop dataframe, roofit and tmva-sofieparser on EPEL 8 ppc64le due to
  "pure virtual method called" errors
Split the root-geom sub-package into three separate sub-packages:
  root-geom, root-geom-builder and root-geom-painter
Enable uring support in EPEL 9 (liburing now available)
New sub-packages: root-geom-webviewer, root-roofit-jsoninterface,
  root-testsupport, root-tree-ntuple-utils, root-tree-webviewer,
  root-xroofit
Dropped patches: 31
New patches: 17
Updated patches: 4
2023-03-19 07:53:00 +01:00

239 lines
8.7 KiB
Diff
Raw Blame History

From 54a05a995bef827413d69b63559494a78ed56946 Mon Sep 17 00:00:00 2001
From: moneta <lorenzo.moneta@cern.ch>
Date: Tue, 14 Feb 2023 11:00:44 +0100
Subject: [PATCH] [tmva] Speed up TMVA CNN and RNN tutorials
Run tutorials with a amximum of 4 threads to avoid MT problems on some machines.
Disable also OpenMP when running in MT in ROOT
Reduce also by a factor of 5 the number of input events
---
tutorials/tmva/TMVA_CNN_Classification.C | 25 +++++++++++++++-------
tutorials/tmva/TMVA_CNN_Classification.py | 12 +++++------
tutorials/tmva/TMVA_RNN_Classification.C | 11 +++++-----
tutorials/tmva/TMVA_RNN_Classification.py | 26 ++++++++++-------------
4 files changed, 39 insertions(+), 35 deletions(-)
diff --git a/tutorials/tmva/TMVA_CNN_Classification.C b/tutorials/tmva/TMVA_CNN_Classification.C
index 0966017ae2..cc6829a24d 100644
--- a/tutorials/tmva/TMVA_CNN_Classification.C
+++ b/tutorials/tmva/TMVA_CNN_Classification.C
@@ -107,7 +107,16 @@ void MakeImagesTree(int n, int nh, int nw)
f.Close();
}
-void TMVA_CNN_Classification(std::vector<bool> opt = {1, 1, 1, 1, 1})
+/// @brief Run the TMVA CNN Classification example
+/// @param nevts : number of signal/background events. Use by default a low value (1000)
+/// but increase to at least 5000 to get a good result
+/// @param opt : vector of bool with method used (default all on if available). The order is:
+/// - TMVA CNN
+/// - Keras CNN
+/// - TMVA DNN
+/// - TMVA BDT
+/// - PyTorch CNN
+void TMVA_CNN_Classification(int nevts = 1000, std::vector<bool> opt = {1, 1, 1, 1, 1})
{
bool useTMVACNN = (opt.size() > 0) ? opt[0] : false;
@@ -125,17 +134,17 @@ void TMVA_CNN_Classification(std::vector<bool> opt = {1, 1, 1, 1, 1})
bool writeOutputFile = true;
- int num_threads = 0; // use default threads
+ int num_threads = 4; // use by default 4 threads if value is not set before
+ // switch off MT in OpenBLAS to avoid conflict with tbb
+ gSystem->Setenv("OMP_NUM_THREADS", "1");
TMVA::Tools::Instance();
// do enable MT running
if (num_threads >= 0) {
ROOT::EnableImplicitMT(num_threads);
- if (num_threads > 0) gSystem->Setenv("OMP_NUM_THREADS", TString::Format("%d",num_threads));
}
- else
- gSystem->Setenv("OMP_NUM_THREADS", "1");
+
std::cout << "Running with nthreads = " << ROOT::GetThreadPoolSize() << std::endl;
@@ -161,7 +170,7 @@ void TMVA_CNN_Classification(std::vector<bool> opt = {1, 1, 1, 1, 1})
The factory is the major TMVA object you have to interact with. Here is the list of parameters you need to pass
- The first argument is the base of the name of all the output
- weightfiles in the directory weight/ that will be created with the
+ weight files in the directory weight/ that will be created with the
method parameters
- The second argument is the output file for the training results
@@ -208,7 +217,7 @@ void TMVA_CNN_Classification(std::vector<bool> opt = {1, 1, 1, 1, 1})
// if file does not exists create it
if (!fileExist) {
- MakeImagesTree(5000, 16, 16);
+ MakeImagesTree(nevts, 16, 16);
}
// TString inputFileName = "tmva_class_example.root";
@@ -289,7 +298,7 @@ void TMVA_CNN_Classification(std::vector<bool> opt = {1, 1, 1, 1, 1})
// Boosted Decision Trees
if (useTMVABDT) {
factory.BookMethod(loader, TMVA::Types::kBDT, "BDT",
- "!V:NTrees=400:MinNodeSize=2.5%:MaxDepth=2:BoostType=AdaBoost:AdaBoostBeta=0.5:"
+ "!V:NTrees=200:MinNodeSize=2.5%:MaxDepth=2:BoostType=AdaBoost:AdaBoostBeta=0.5:"
"UseBaggedBoost:BaggedSampleFraction=0.5:SeparationType=GiniIndex:nCuts=20");
}
/**
diff --git a/tutorials/tmva/TMVA_CNN_Classification.py b/tutorials/tmva/TMVA_CNN_Classification.py
index 0760bacfd9..85fde20763 100644
--- a/tutorials/tmva/TMVA_CNN_Classification.py
+++ b/tutorials/tmva/TMVA_CNN_Classification.py
@@ -25,6 +25,9 @@
import ROOT
+#switch off MT in OpenMP (BLAS)
+ROOT.gSystem.Setenv("OMP_NUM_THREADS", "1")
+
TMVA = ROOT.TMVA
TFile = ROOT.TFile
@@ -105,6 +108,7 @@ def MakeImagesTree(n, nh, nw):
hasGPU = ROOT.gSystem.GetFromPipe("root-config --has-tmva-gpu") == "yes"
hasCPU = ROOT.gSystem.GetFromPipe("root-config --has-tmva-cpu") == "yes"
+nevt = 1000 # use a larger value to get better results
opt = [1, 1, 1, 1, 1]
useTMVACNN = opt[0] if len(opt) > 0 else False
useKerasCNN = opt[1] if len(opt) > 1 else False
@@ -141,17 +145,13 @@ if not useTMVACNN:
writeOutputFile = True
-num_threads = 0 # use default threads
+num_threads = 4 # use default threads
max_epochs = 10 # maximum number of epochs used for training
# do enable MT running
if num_threads >= 0:
ROOT.EnableImplicitMT(num_threads)
- if (num_threads > 0) :
- ROOT.gSystem.Setenv("OMP_NUM_THREADS", str(num_threads))
-else:
- ROOT.gSystem.Setenv("OMP_NUM_THREADS", "1")
print("Running with nthreads = ", ROOT.GetThreadPoolSize())
@@ -218,7 +218,7 @@ inputFileName = "images_data_16x16.root"
# if the input file does not exist create it
if ROOT.gSystem.AccessPathName(inputFileName):
- MakeImagesTree(5000, 16, 16)
+ MakeImagesTree(nevt, 16, 16)
inputFile = TFile.Open(inputFileName)
if inputFile is None:
diff --git a/tutorials/tmva/TMVA_RNN_Classification.C b/tutorials/tmva/TMVA_RNN_Classification.C
index 2711673ec9..8701ba6c19 100644
--- a/tutorials/tmva/TMVA_RNN_Classification.C
+++ b/tutorials/tmva/TMVA_RNN_Classification.C
@@ -136,13 +136,14 @@ void MakeTimeData(int n, int ntime, int ndim )
}
}
/// macro for performing a classification using a Recurrent Neural Network
+/// @param nevts = 2000 Number of events used. (increase for better classification results)
/// @param use_type
/// use_type = 0 use Simple RNN network
/// use_type = 1 use LSTM network
/// use_type = 2 use GRU
/// use_type = 3 build 3 different networks with RNN, LSTM and GRU
-void TMVA_RNN_Classification(int use_type = 1)
+void TMVA_RNN_Classification(int nevts = 2000, int use_type = 1)
{
const int ninput = 30;
@@ -150,7 +151,7 @@ void TMVA_RNN_Classification(int use_type = 1)
const int batchSize = 100;
const int maxepochs = 20;
- int nTotEvts = 10000; // total events to be generated for signal or background
+ int nTotEvts = nevts; // total events to be generated for signal or background
bool useKeras = true;
@@ -190,14 +191,12 @@ void TMVA_RNN_Classification(int use_type = 1)
useKeras = false;
#endif
- int num_threads = 0; // use by default all threads
+ int num_threads = 4; // use by default all threads
+ gSystem->Setenv("OMP_NUM_THREADS", "1"); // switch off MT in OpenBLAS
// do enable MT running
if (num_threads >= 0) {
ROOT::EnableImplicitMT(num_threads);
- if (num_threads > 0) gSystem->Setenv("OMP_NUM_THREADS", TString::Format("%d",num_threads));
}
- else
- gSystem->Setenv("OMP_NUM_THREADS", "1");
TMVA::Config::Instance();
diff --git a/tutorials/tmva/TMVA_RNN_Classification.py b/tutorials/tmva/TMVA_RNN_Classification.py
index 88280a5849..625a120b3d 100644
--- a/tutorials/tmva/TMVA_RNN_Classification.py
+++ b/tutorials/tmva/TMVA_RNN_Classification.py
@@ -22,6 +22,16 @@
import ROOT
+num_threads = 4 # use max 4 threads
+# do enable MT running
+if ROOT.gSystem.GetFromPipe("root-config --has-imt") == "yes":
+ ROOT.EnableImplicitMT(num_threads)
+ ROOT.gSystem.Setenv("OMP_NUM_THREADS", "1") # switch OFF MT in OpenBLAS
+ print("Running with nthreads = {}".format(ROOT.GetThreadPoolSize()))
+else:
+ print("Running in serail mode since ROOT does not support MT")
+
+
TMVA = ROOT.TMVA
TFile = ROOT.TFile
@@ -141,7 +151,7 @@ ntime = 10
batchSize = 100
maxepochs = 10
-nTotEvts = 10000 # total events to be generated for signal or background
+nTotEvts = 2000 # total events to be generated for signal or background
useKeras = True
@@ -184,22 +194,8 @@ if ROOT.gSystem.GetFromPipe("root-config --has-tmva-pymva") == "yes":
else:
useKeras = False
-num_threads = 0 # use by default all threads
-# do enable MT running
-if ROOT.gSystem.GetFromPipe("root-config --has-imt") == "yes":
- if num_threads >= 0:
- ROOT.EnableImplicitMT(num_threads)
- if num_threads > 0:
- ROOT.gSystem.Setenv("OMP_NUM_THREADS", str(num_threads))
- else:
- ROOT.gSystem.Setenv("OMP_NUM_THREADS", "1")
- print("Running with nthreads = {}".format(ROOT.GetThreadPoolSize()))
-
-else:
- print("Running in serail mode since ROOT does not support MT")
-
inputFileName = "time_data_t10_d30.root"
fileDoesNotExist = ROOT.gSystem.AccessPathName(inputFileName)
--
2.39.2
Reference in a new issue View git blame Copy permalink