add wrapper for training Spec2Vec in Galaxy

[maximskorik]

Updates

• add spec2vec_training tool to Galaxy

Description

This tool can train a Spec2Vec model from msp or mgf spectra files. The outputs are json model-metadata and npy peak-embeddings, which can be read by spec2vec.serialization module (in the future downstream step) to compute Spec2Vec similarity scores. Optional outputs are Python pickle file & checkpoints of a model at user-defined iterations.

Closes #315

[xtrojak]

@maximskorik can you please explain the role of docker image? Is it just to get the newest version of the tool available in master branch?

[maximskorik]

@maximskorik can you please explain the role of docker image? Is it just to get the newest version of the tool available in master branch?

Exactly. We recently added functionality for Spec2Vec to export a model to json+npy files and it hasn't been released yet. Hence the docker image with a tag that points to the merge commit with this update until these changes are released and become available via Conda.

[maximskorik]

The CI fails because the linter is concerned about suffixes in the value and delta attributes of <has_size>. I've submitted an issue to Planemo galaxyproject/planemo#1335.

[xtrojak]

Looks pretty good! Great job @maximskorik

[hechth]

Can we maybe change this to be compliant with the CI so that it passes? I know it is not a fault on our side, but still as soon as the CI fails and you merge stuff, it becomes harder to see when it actually starts failing for a different reason.

[maximskorik]

I removed suffixes from the size attributes in tests (70991af) since planemo v0.75.3 wouldn't accept them anymore, causing the tests to fail. The CI now passes.

cc @hechth