CLI

Command line interface for CAMPA.

Use this to execute long-running steps of the CAMPA workflow (e.g. clustering large data or calculating co-occurrence features) in a script on e.g. a HPC system.

See the Workflow documentation for a short introduction to the CLI commands and the Tutorials for the CLI commands needed to reproduce the example workflow.

Basic usage

CLI for CAMPA - conditional autoencoder for multiplexed image analysis.

usage: campa <command> [<args>]

Available subcommands are:
setup               Create configuration file ``campa.ini``
create_dataset      Create NNDataset using parameter file
train               Train and evaluate models defined by experiment config
cluster             Cluster data and project clustering
extract_features    Extract features from clustered dataset

Positional Arguments

command: Subcommand to run.

campa setup

Create configuration file campa.ini.

usage: campa setup [-h] [--quiet]

Named Arguments

--quiet

create default configuration file without asking for user input.

Default: False

campa create_dataset

Create NNDataset using parameter file

usage: campa create_dataset [-h] params

Positional Arguments

params: path to data_params.py

campa train

Train and evaluate models defined by experiment config

usage: campa train [-h] (--config CONFIG | --experiment-dir EXPERIMENT_DIR)
                   [--exp-name [EXP_NAME ...]]
                   {all,train,evaluate,trainval,compare}

Positional Arguments

mode

Possible choices: all, train, evaluate, trainval, compare

Default: “all”

Named Arguments

--config: path to experiment_config.py
--experiment-dir: experiment_dir containing experiment folders. Relative to EXPERIMENT_DIR
--exp-name: Select names of experiments to run. If not specified, all available experiments are run

campa cluster

Cluster data and project clustering.

usage: campa cluster [-h] experiment-dir {create,prepare-full,project} ...

Positional Arguments

experiment-dir: relative to EXPERIMENT_PATH

Sub-commands

create

Create (subsampled) data for clustering. Uses all data used to train exp

campa cluster create [-h] [--subsample] [--frac FRAC] [--save-dir SAVE_DIR]
                     [--cluster]

Named Arguments

--subsample

Subsample the data

Default: False

--frac

Fraction of pixels to use for clustering

Default: 0.005

--save-dir

directory to save subsampled cluster data, relative to experiment-dir. Default is aggregated/sub-FRAC

--cluster

use cluster parameters in Experiment config to cluster the subsetted data.

Default: False

prepare-full

Prepare full data for clustering. Predicts cluster-rep.

campa cluster prepare-full [-h] [--save-dir SAVE_DIR]
                           [--data-dirs DATA_DIRS [DATA_DIRS ...]]

Named Arguments

--save-dir

directory to save prepared full data to, relative to experiment-dir.

Default: “aggregated/full_data”

--data-dirs

Data directories to prepare. Defaults to experiment data directories

project

Project existing clustering.

campa cluster project [-h] [--save-dir SAVE_DIR] [--data-dir DATA_DIR]
                      [--cluster-name CLUSTER_NAME]
                      cluster-data-dir

Positional Arguments

cluster-data-dir: directory in which clustering is stored relative to experiment-dir. Usually in aggregated/sub-FRAC

Named Arguments

--save-dir

directory in which data to be projected is stored, relative to experiment-dir.

Default: “aggregated/full_data”

--data-dir

data to project. If not specified, project all data_dirs in save_dir

--cluster-name

name of clustering to project

Default: “clustering”

campa extract_features

Extract features from clustered dataset.

usage: campa extract_features [-h] params

Positional Arguments

params: path to feature_params.py