Command builder · cryoDRGN

Pick a cryoDRGN command. The line below lists required arguments plus options you actually set (empty fields and CLI defaults are omitted).

Required

Particles

Output folder (-o)

--zdim

--ctf

Dataset
loading

--relion31relion 3.1 data format --uninvert-datanot inverting input data (e.g. for empiar-10076) --lazylazy data loading

--indfiltering .pkl

--datadir

--max-threads

Training
parameters

On by default — uncheck for --no-shuffle --multigpu Mixed precision (AMP) — uncheck for --no-amp

-n totalepochs

--epochs-pose-search epochs

--n-imgs-pose-searchimages to train for pose search

--epochs-sgd epochs

--pose-only-phase epochs

--num-workerssubprocesses to use for data loading

--shuffler-size

All batch sizes

--batch-size-hps images

--batch-size-known-poses images

--batch-size-sgd images

Advanced parameters

--loadcheckpoint

--load-posespose file

--seed

-v

--log-intervallogging interval in n_imgs

--checkpointcheckpointing interval in n_epochs

--verbose-time

--lrlearning rate (optimizer)

--lr-pose-tablelearning rate (pose table)

--lr-conf-tablelearning rate (conf table)

--lr-conf-encoderlearning rate (conf encoder)

--wdweight decay

--hypervolume-optimizer-typeoptimizer type

adam

--pose-table-optimizer-typeoptimizer (pose table)

adam lbfgs

--conf-table-optimizer-typeoptimizer (conformation table)

adam lbfgs

--conf-encoder-optimizer-typeoptimizer (conformation encoder)

adam

--output-masktype of output mask to use

circ frequency_marching

--add-one-frequency-every

--n-frequencies-per-epoch epochs

--max-freq

--window-radius-gt-real

--beta-conf

--trans-l1-regularizer

--l2-smoothness-regularizer

--variational-het

--std-z-init

--use-conf-encoder

--depth-cnndepth of the encoder

--channels-cnnchannels in the encoder

--kernel-size-cnnsize of the kernels in the encoder

--resolution-encoder

--explicit-volumeuse an explicit volume (voxel array)

--layershidden layers in the hypervolume

--dimhidden layers in the hypervolume

--pe-type

gaussian

--pe-dimfrequencies used for positional encoding

--feat-sigma

--hypervolume-domaindomain of the hypervolume

hartley

--pe-type-conf

(default None) geom

--initial-confinitial conformations

--n-imgs-pretrainimages to use for pre-training

--l-start

--l-end

--niterpose search iterations

--t-extentextent of the translation search grid

--t-ngrid

--t-xshift

--t-yshift

--no-trans-search-at-pose-search

--nkeptposesposes kept per image

--base-healpybase healpy index

--no-trans

--normdata normalization as shift (shift, scale)

On by default — uncheck for --no-analysis

Required

Particles

Output folder (-o)

--poses

--zdim

--ctf

Dataset
loading

--uninvert-data On by default — uncheck for --no-window --lazy

--indfiltering .pkl

--datadir

--window-rwindowing radius

--shuffler-sizeif non-zero

--num-workers

--max-threads

Training
parameters

Mixed precision (AMP) — uncheck for --no-amp --multigpu

-n totalepochs

-bminibatch size images

--wdweight decay in adam optimizer

--lrlearning rate in adam optimizer

--beta

--beta-control

--normdata normalization as shift (shift, scale)

Advanced parameters

--loadinitialize training from a checkpoint

On by default — uncheck for --no-analysis

--checkpointcheckpointing interval in n_epochs

--log-intervallogging interval in n_imgs

-v

--seedrandom seed

--shuffle-seedrandom seed for data shuffling

--do-pose-sgdrefine poses with gradient descent

--pretrainepochs with fixed poses before pose sgd

--emb-typeso(3) embedding type for pose sgd

s2s2 quat

--pose-lrlearning rate for pose optimizer

--enc-layershidden layers

--enc-dimnodes in hidden layers

--encode-modetype of encoder network

conv resid mlp tilt

--enc-maskcircular mask of image for encoder

--use-real

--dec-layershidden layers

--dec-dimnodes in hidden layers

--pe-typetype of positional encoding

geom_ft geom_full geom_lowf geom_nohighf linear_lowf gaussian none

--feat-sigmascale for random gaussian features

--pe-dimnum frequencies in positional encoding

--domainvolume decoder representation

hartley fourier

--activationactivation

relu leaky_relu

Required

Particles

Output folder (-o)

--poses

--ctf

Dataset
loading

--uninvert-data On by default — uncheck for --no-window --lazy

--indfiltering .pkl

--datadir

--window-rwindowing radius

--shuffler-sizeif non-zero

Training
parameters

Mixed precision (AMP) — uncheck for --no-amp --multigpu

-n totalepochs

-bminibatch size images

--wdweight decay in adam optimizer

--lrlearning rate in adam optimizer

--normdata normalization as shift (shift, scale)

Advanced parameters

--loadinitialize training from a checkpoint

--checkpointcheckpointing interval in n_epochs

--log-intervallogging interval in n_imgs

-v

--seedrandom seed

--shuffle-seedrandom seed for data shuffling

--do-pose-sgdrefine poses with gradient descent

--pretrainepochs with fixed poses before pose sgd

--emb-typeso(3) embedding type for pose sgd

s2s2 quat

--pose-lrlearning rate for pose optimizer

--layershidden layers

--dimnodes in hidden layers

--l-extent

--pe-typetype of positional encoding

geom_ft geom_full geom_lowf geom_nohighf linear_lowf gaussian none

--pe-dimnum frequencies in positional encoding

--domainvolume decoder representation

hartley fourier

--activationactivation

relu leaky_relu

--feat-sigmascale for random gaussian features

Required

Particles

Output folder (-o)

--poses

--zdim

--ctf

Dataset
loading

--uninvert-data On by default — uncheck for --no-window --lazy

--indfiltering .pkl

--datadir

--window-rwindowing radius

--shuffler-sizeif non-zero

Training
parameters

Mixed precision (AMP) — uncheck for --no-amp --multigpu

-n totalepochs

-bminibatch size images

--wdweight decay in adam optimizer

--lrlearning rate in adam optimizer

--normdata normalization as shift (shift, scale)

Advanced parameters

--loadinitialize training from a checkpoint

On by default — uncheck for --no-analysis

--checkpointcheckpointing interval in n_epochs

--log-intervallogging interval in n_imgs

-v

--seedrandom seed

--shuffle-seedrandom seed for data shuffling

--load-z

--z-lrlearning rate for latent z optimizer

--pretrain-zepochs to pretrain z before training model

--do-pose-sgdrefine poses with gradient descent

--pretrain-poseepochs with fixed poses before pose sgd

--emb-typeso(3) embedding type for pose sgd

s2s2 quat

--pose-lrlearning rate for pose optimizer

--layershidden layers

--dimnodes in hidden layers

--l-extent

--pe-typetype of positional encoding

geom_ft geom_full geom_lowf geom_nohighf linear_lowf gaussian none

--pe-dimnum frequencies in positional encoding

--domainvolume decoder representation

hartley fourier

--activationactivation

relu leaky_relu

--feat-sigmascale for random gaussian features

Required

Particles

Output folder (-o)

--poses

--ctf

Dataset
loading

--uninvert-data --lazy

--indfiltering .pkl

--datadir

--first

Backprojection
parameters

On by default — uncheck for --no-half-maps On by default — uncheck for --no-fsc-vals --output-sumcount

-bimages to process in each batch images

--ctf-alg

flip mul

--reg-weight

--log-intervallogging interval in n_imgs

Tilt series parameters

Tomography tilt geometry and per-tilt treatments.

--tilt --force-ntilts

--ntiltstilts per particle to backproject

-d

-a

Advanced parameters

Required

Particles

Output folder (-o)

--zdim

--ctf

Dataset
loading

--uninvert-data On by default — uncheck for --no-window --lazy

--indfiltering .pkl

--datadir

--window-rwindowing radius

--shuffler-sizeif non-zero

--max-threads

Training
parameters

Mixed precision (AMP) — uncheck for --no-amp --multigpu

-n totalepochs

-bminibatch size images

--wdweight decay in adam optimizer

--lrlearning rate in adam optimizer

--beta

--beta-control

--normdata normalization as shift (shift, scale)

--equivariancestrength of equivariance loss

--eq-start-itit at which equivariance turned on

--eq-end-itit at which equivariance max

--l-ramp-epochs epochs

--l-ramp-modelif 1

--reset-model-everyif set

--reset-optim-everyif set

--reset-optim-after-pretrainif set

Advanced parameters

--loadinitialize training from a checkpoint

--load-posesinitialize training from a checkpoint

On by default — uncheck for --no-analysis

--checkpointcheckpointing interval in n_epochs

--log-intervallogging interval in n_imgs

-v

--seedrandom seed

--shuffle-seedrandom seed for data shuffling

--enc-only

--l-startstarting l radius

--l-endend l radius

--niteriterations of grid subdivision

--t-extent+/- pixels to search over translations

--t-ngridinitial grid size for translations

--t-xshiftx-axis translation shift

--t-yshifty-axis translation shift

--pretraininitial iterations with random poses

--ps-freqfrequency of pose inference

--nkeptposes

--base-healpy

--pose-model-update-freqif set, only update the model

--enc-layershidden layers

--enc-dimnodes in hidden layers

--encode-modetype of encoder network

conv resid mlp tilt

--enc-maskcircular mask of image for encoder

--use-real

--dec-layershidden layers

--dec-dimnodes in hidden layers

--pe-typetype of positional encoding

geom_ft geom_full geom_lowf geom_nohighf linear_lowf gaussian none

--feat-sigmascale for random gaussian features

--pe-dimnum frequencies in positional encoding

--domainvolume decoder representation

hartley fourier

--activationactivation

relu leaky_relu

Required

Particles

Output folder (-o)

--ctf

Dataset
loading

--uninvert-data On by default — uncheck for --no-window --lazy

--indfiltering .pkl

--datadir

--window-rwindowing radius

--shuffler-sizeif non-zero

Training
parameters

--no-transdon't search over translations Mixed precision (AMP) — uncheck for --no-amp

--t-extent+/- pixels to search over translations

--t-ngridinitial grid size for translations

--t-xshiftx-axis translation shift

--t-yshifty-axis translation shift

--pretraininitial iterations with random poses

--ps-freqfrequency of pose inference

-n totalepochs

-bminibatch size images

--wdweight decay in adam optimizer

--lrlearning rate in adam optimizer

--normdata normalization as shift (shift, scale)

Advanced parameters

--loadinitialize training from a checkpoint

--load-posesinitialize training from a checkpoint

--checkpointcheckpointing interval in n_epochs

--log-intervallogging interval in n_imgs

-v

--seedrandom seed

--shuffle-seedrandom seed for data shuffling

--tiltparticle stack file (.mrcs)

--tilt-degx-axis tilt offset in degrees

--l-startstarting l radius

--l-endend l radius

--niteriterations of grid subdivision

--l-ramp-epochs epochs

--probabilisticuse probabilistic bound

--nkeptposes

--base-healpy

--pose-model-update-freqif set, only update the model

--layershidden layers

--dimnodes in hidden layers

--l-extent

--pe-typetype of positional encoding

geom_ft geom_full geom_lowf geom_nohighf linear_lowf gaussian none

--pe-dimnum frequencies in positional encoding

--domainvolume decoder representation

hartley fourier

--activationactivation

relu leaky_relu

--feat-sigmascale for random gaussian features

Required

Workdir

Epoch epochs

Output folder (-o)

Run
options

--skip-volskip generation of volumes --skip-umapskip running umap

--deviceoptionally specify cuda device

Volumes
to generate

--pcprincipal component traversals to generate

--n-per-pc

--ksamplekmeans samples to generate

Advanced parameters

--Apix

--flipflip handedness of output volumes

--invertinvert contrast of output volumes

-d

--low-pass

--crop

--vol-start-index

Required

Workdir

Epoch epochs

Output folder (-o)

Run
options

--multigpu --skip-umapskip running umap

--deviceoptionally specify cuda device

--vol-indindex .pkl for filtering volumes

Extra
arguments for volume generation

--flipflip handedness of output volume --skip-volskip generation of volumes

-N totalepochs

--Apixpixel size to add to .mrc header

-d

--vol-start-index

Advanced parameters

--threshdensity value to threshold for masking

--dilatedilate initial mask by this amount

--cosine-edgeapply a cosine edge to the mask

--maskcustom mask

--linkage

-Mclusters

--pc-dimpca dimensionality reduction

--plot-dimdimensions to plot

Required

Workdir

Epoch epochs

Output folder (-o)

Run
options

--deviceoptionally specify cuda device

--landscape-dirlandscape analysis directory

--seedrandom seed

Volume
generation arguments

--flipflip handedness --skip-volskip generation of volumes

-N totalepochs

-d

Advanced parameters

All batch sizes

--batch-size images

--test-batch-size images

--epochs epochs

--lrlearning rate

--dimmlp hidden layer dimension

--layersmlp number of hidden layers

--num-neighborsnearest neighbors to consider

--resolutionleiden resolution

cryodrgn abinit …

This page does not run commands — copy the line into your shell or job script.

Datasetloading

Trainingparameters

Advanced parameters

Datasetloading

Trainingparameters

Advanced parameters

Datasetloading

Trainingparameters

Advanced parameters

Datasetloading

Trainingparameters

Advanced parameters

Datasetloading

Backprojectionparameters

Tilt series parameters

Advanced parameters

Datasetloading

Trainingparameters

Advanced parameters

Datasetloading

Trainingparameters

Advanced parameters

Runoptions

Volumesto generate

Advanced parameters

Runoptions

Extraarguments for volume generation

Advanced parameters

Runoptions

Volumegeneration arguments

Advanced parameters

Dataset
loading

Training
parameters

Dataset
loading

Training
parameters

Dataset
loading

Training
parameters

Dataset
loading

Training
parameters

Dataset
loading

Backprojection
parameters

Dataset
loading

Training
parameters

Dataset
loading

Training
parameters

Run
options

Volumes
to generate

Run
options

Extra
arguments for volume generation

Run
options

Volume
generation arguments