dl4jets/deepjet

DeepJet: Repository for training and evaluation of deep neural networks for Jet identification

This package depends on DeepJetCore 3.X (https://github.com/DL4Jets/DeepJetCore). (but it should still work for DJC2)

Usage

After logging in and setting up the DeepJetCore singularity environment, please source the DeepJet environment (please cd to the directory first!).

cd <your working dir>/DeepJet
source env.sh

The preparation for the training consists of the following steps

• define the data structure for the training. The DeepJet datastructure is found in the modules directory as the class TrainData_DF. DeepCSV is found as the class TrainData_DeepCSV

• convert the root file to the data strucure for training using DeepJetCore tools. The class argument should be TrainData_DF for DeepJet and TrainData_DeepCSV for DeepCSV:

    convertFromSource.py -i /path/to/the/root/ntuple/list_of_root_files.txt -o /output/path/that/needs/some/disk/space -c TrainData_DF
  

  This step can take a while.

• prepare the training file and the model. Please refer to DeepJet/Train/train_DeepFlavour.py for training of DeepJet and DeepJet/Train/train_DeepCSV.py for DeepCSV.

Training

Since the training can take a while, it is advised to open a screen session, such that it does not die at logout.

ssh lxplus.cern.ch
<note the machine you are on, e.g. lxplus058>
screen
ssh lxplus7

Then source the environment, and proceed with the training. Detach the screen session with ctr+a d. You can go back to the session by logging in to the machine the session is running on (e.g. lxplus58):

ssh lxplus.cern.ch
ssh lxplus058
screen -r

Please close the session when the training is finished

the training is launched in the following way:

python3 train_DeepFlavour.py /path/to/the/output/of/convert/dataCollection.djcdc <output dir of your choice>

Evaluation

After the training has finished, the performance can be evaluated.

predict.py <output dir of training>/KERAS_model.h5  <output dir of training>/trainsamples.dc <dir with test sample stored as rootfiles>/filelist.txt <output directory>

This creates output trees with the prediction scores as well as truth information and some kinematic variables.

deepjet's Issues

activating deepjetpkg by sourcing lxplus_env.sh / gpu_env.sh

https://github.com/SwapneelM/DeepJet
while trying to follow the instructions for activating the environment using

source lxplus_env.sh / gpu_env.sh

export PATH=$DEEPJETCORE/bin:$PATH
: No such file or directory

but if I try to activate to activate using
source pypkg_env.sh
the deepjetpkg activates with no issue

Is this something which can be an issue while training?

Issues with new uproot version

The Conversion script in the DeepJetCore-repository fails with a TypeError when executed with the TrainData_DeepCSV module and very likely also with the DeepFlavour module.

The stack trace:

Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
    self.run()
  File "/usr/lib/python3.6/multiprocessing/process.py", line 93, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/share/DJC/DeepJetCore/DataCollection.py", line 427, in writeData_async
    td.writeFromSourceFile(tmpinput, self.weighterobjects, istraining = not self.istestdata, outname=newpath)
  File "/usr/share/DJC/DeepJetCore/TrainData.py", line 105, in writeFromSourceFile
    self.readFromSourceFile(filename, weighterobjects, istraining)
  File "/usr/share/DJC/DeepJetCore/TrainData.py", line 81, in readFromSourceFile
    x,y,w = self.convertFromSourceFile(filename, weighterobjects, istraining,  **kwargs)
  File "/afs/cern.ch/work/n/neich/private/BTV-HLT-training-tools/DeepJet/modules/datastructures/TrainData_deepFlavour.py", line 448, in convertFromSourceFile
    truth = reduceTruth(truth_arrays)
  File "/afs/cern.ch/work/n/neich/private/BTV-HLT-training-tools/DeepJet/modules/datastructures/TrainData_deepFlavour.py", line 428, in reduceTruth
    return np.vstack((b+lepb,bb+gbb,c+cc+gcc,uds+g)).transpose()
  File "<__array_function__ internals>", line 6, in vstack
TypeError: no implementation found for 'numpy.vstack' on types that implement __array_function__: [<class 'awkward.highlevel.Array'>]

This issue is not present, when using uproot version 3 instead of the newest version uproot 4.

A possible hotfix for that could be changing imports

import uproot
...

to

import uproot3 as uproot
...

For the module.

module object not callable while tring to convert root file to training format

https://github.com/SwapneelM/DeepJet/blob/master/README.md
the instructions for preprocessing the root file into training format
convertFromRoot.py -i /path/to/the/root/ntuple/list_of_root_files.txt -o /output/path/that/needs/some/disk/space -c TrainData_myclass

I used the following command
convertFromRoot.py -i /eos/user/e/ebols/HLT/DeepNTuple_1p1MJets_changedDeepCSVVtxDef.root -o /afs/cern.ch/work/r/runiyal/DeepJet -c TrainDataDeepJet

which gave me the following error

(deepjetpkg) [runiyal@lxplus027 DeepJet]$ convertFromRoot.py -i /eos/user/e/ebols/HLT/DeepNTuple_1p1MJets_changedDeepCSVVtxDef.root -o /afs/cern.ch/work/r/runiyal/DeepJet -c TrainDataDeepJet
Using TensorFlow backend.
INFO:root:infile = /eos/user/e/ebols/HLT/DeepNTuple_1p1MJets_changedDeepCSVVtxDef.root
INFO:root:outPath = /afs/cern.ch/work/r/runiyal/DeepJet
Traceback (most recent call last):
File "/afs/cern.ch/work/r/runiyal/miniconda3/envs/deepjetpkg/bin/convertFromRoot.py", line 4, in
import('pkg_resources').run_script('DeepJetCore==1.0.0', 'convertFromRoot.py')
File "/afs/cern.ch/work/r/runiyal/miniconda3/envs/deepjetpkg/lib/python2.7/site-packages/pkg_resources/init.py", line 658, in run_script
self.require(requires)[0].run_script(script_name, ns)
File "/afs/cern.ch/work/r/runiyal/miniconda3/envs/deepjetpkg/lib/python2.7/site-packages/pkg_resources/init.py", line 1438, in run_script
exec(code, namespace, namespace)
File "/afs/cern.ch/work/r/runiyal/miniconda3/envs/deepjetpkg/lib/python2.7/site-packages/DeepJetCore-1.0.0-py2.7-linux-x86_64.egg/EGG-INFO/scripts/convertFromRoot.py", line 118, in
infile, traind(class_args) if class_args else traind(), outPath,
TypeError: 'module' object is not callable