calc.descriptor.fingerprints accumulates the previous fingerprint existing in database

[Javad Hashemi]

We have 100 images in each of "training-2000k.traj" and "test-2300k.traj". If in ipython we run the following code:

import amp
from amp import Amp
from amp.descriptor.gaussian import Gaussian
from amp.model.neuralnetwork import NeuralNetwork

images = amp.hash_images('training-2000k.traj')
calc = Amp(descriptor=Gaussian(), model=NeuralNetwork())
calc.descriptor.calculate_fingerprints(images)
trainFingerPrints = calc.descriptor.fingerprints
trainFingerPrints.open()

then the number of images loaded is 100 as it should be:

In [2]: len(trainFingerPrints.d.keys())
Out[2]: 100

then if we change the name of the file from 'training-2000k.traj' to "test-2300k.traj" and run the code again without removing .ampdb folder we will have

In [4]: len(trainFingerPrints.d.keys())
Out[4]: 199

Although we should just get 100

[Muammar El Khatib]

The two trajectory files don't have the same images present and therefore this should be the expected behavior.