Cli Interface

The cli Interface can be accessed globaly after running julia --project build/build.jl install.

	~/.julia/bin/mlnanoshaper train [options] [flags]

Train a model.

Intro

Train a model that can reconstruct a protein surface using Machine Learning. Default value of parameters are specified in the param/param.toml file. In order to override the param, you can use the differents options.

Options

• --nb-epoch <Int>: the number of epoch to compute.
• --model, -m <String>: the model name. Can be anakin.
• --nb-data-points <Int>: the number of proteins in the dataset to use
• --name, -n <String>: name of the training run
• --cutoff-radius, -c <Float32>: the cutoff_radius used in training
• --ref-distance <Float32>: the reference distane (in A) used to rescale distance to surface in loss
• --learning-rate, -l <Float64>: the learning rate use by the model in training.
• --loss <String>: the loss function, one of "categorical" or "continuous".

Flags

• --gpu, -g: should we do the training on the gpu, does nothing currently.