Spectromer

A pretrained spectra transformer.

Installation

As usual, you can install Spectromer using pip:

ACTUALLY IT IS NOT YET PUBLISHED ON PYPI, WILL DO SOON

pip install spectromer

Training Spectromer

To train Spectromer, you can use the spectromer train command. The following command trains Spectromer on the GNPS dataset with 256 peaks, 256 embedding size, 1000 epochs, 4096 batch size, and saves the model to v1.keras.

spectromer train --dataset GNPS --number-of-peaks 256 --embedding-size 512 --epochs 10000 --batch-size 1024 --output v1.keras --verbose

Embedding spectra

To embed spectra, you can use the spectromer embed command. The following command embeds the spectra in spectra.mgf using the model v1.keras and saves the embeddings to embeddings.npy.

spectromer transform --input datasets/GNPS/matchms.mgf --model v1.keras --output matchms.npy --verbose