tomodrgn convergence_nn#

Purpose#

Evaluate similarity of a reference reconstruction to each reconstruction generated by train_nn at specified epochs throughout training, calculated via map-map FSC.

Sample usage#

The examples below are adapted from tomodrgn/testing/commandtest*.py, and rely on other outputs from commandtest.py to execute successfully.

# Warp v1 style inputs
tomodrgn \
    convergence_nn \
    output/nn_classE_sim_dosetiltweightmask \
    data/10076_classE_32.mrc \
    --fsc-mask soft

# WarpTools style inputs
tomodrgn \
    convergence_nn \
    output/nn_warptools_70S_dosetiltweightmask \
    data/warptools_test_box-32_angpix-12_reconstruct.mrc \
    --fsc-mask soft

Arguments#

usage: convergence_nn [-h] [--max-epoch MAX_EPOCH] [--include-dc]
                      [--fsc-mask {none,sphere,tight,soft}]
                      [--plot-format {png,svgz}]
                      training_directory reference_volume

Positional Arguments#

training_directory

train_nn directory containing reconstruct.N.mrc

reference_volume

volume against which to calculate FSC

Named Arguments#

--max-epoch

Maximum epoch for which to calculate FSCs

--include-dc

Include FSC calculation for DC component, default False because DC component default excluded during training

Default: False

--fsc-mask

Possible choices: none, sphere, tight, soft

Type of mask applied to volumes before calculating FSC

Default: 'soft'

--plot-format

Possible choices: png, svgz

File format with which to save plots

Default: 'png'

Common next steps#

  • Extend model training with tomodrgn train_nn [...] --load latest if not yet converged

  • Train a heterogeneous model with tomodrgn train_vae