This is the root directory of the hqsp project
- Modify training for an end-to-end pipeline (generating data is now part of the learning process)
- Adapt qft circuit so that it contains learnable parameters
- Integrate entangling dropout?
- Conduct a few experiments
The training can be executed by running train.py. This includes:
- check versioning for changes
- generating waveforms (spectrograms)
- generating quantum data
- training the network
Waveforms can be loaded from disk by including --waveform=0 as argument.
Quantum data can be loaded from disk by including --quantum=0 as argument.
The Pixel-Channel-Mapping can be activated by setting --quantum=-1.
Checking the versioning directory can be disabled by --checkTree=-1.
Actual training can be disabled by --train=0.
Paths in the script need to be adapted to your needs.
Similar to training:
- check versioning for changes
- generating waveforms (spectrograms)
- generating quantum data
- testing the network
Waveforms can be loaded from disk by including --waveform=0 as argument.
Quantum data can be loaded from disk by including --quantum=0 as argument.
The Pixel-Channel-Mapping can be activated by setting --quantum=-1.
Checking the versioning directory can be disabled by --checkTree=-1.
Paths in the script need to be adapted to your needs.
General training procedure: train.py General testing procedure: test.py Quick evaluations and testing: eval.py Extraction of test data: extractTestData.py Training, loading and evaluation of model: fitModel.py Spectrogram generation, global parameter storage and multiprocessing: generateFeatures.py Experiment viewer: viewer.py