cryodrgn abinit command API

full list of parameters that can be used to control the behavior of the cryodrgn abinit command

Required parameters

particles The input particle stack (.mrcs, .star, .cs, or .txt).

--outdir/-o The directory where model outputs will be saved.

--ctf Path to CTF parameters (.pkl). Must be the size of the full dataset.

--zdim Dimension of conformation z-latent-space — 0 for homogeneous reconstruction

Other input parameters

--ind Path to indices (.pkl) or number of images to keep (first images kept). Default is to use full dataset if not given.

--datadir When using a .star or .cs file with relative paths, path to the directory containing the .mrcs files.

--relion31 Flag for RELION3.1 .star file format.

--uninvert-data Flag for not inverting input data (e.g. for EMPIAR-10076).

--load Path to a checkpoint (weights.<epoch>.pkl) to load.

--load-poses Path to a pose file (pose.<epoch>.pkl) to load.

--lazy Flag for lazy data loading, which uses less GPU memory.

--max-threads Number of threads to use for data loading.

--no-trans Flag for indicating that the dataset does not contain translations.

--norm Override the default method for normalizing the images stack using two values (mean, std).

Training scheduling

--num-epochs Number of total epochs to train for.

--epochs-sgd Number of epochs to train for SGD.

--epochs-pose-search Number of epochs for hierarchical pose search.

--n-imgs-pose-search Alternatively, the number of images to use for hierarchical pose search instead of specifying in epochs.

Loss

--beta-conf [string, default = 0.0] Beta term penalizing the KL divergence of the posterior distribution of conformations. Only used in variational mode.

--trans-l1-regularizer [string, default = 0.0] Strength of the L1 regularizer applied on estimated translations.

--l2-smoothness-regularizer [string, default = 0.0] Strength of the L2 smoothness regularization (penalization of strong gradients).

Optimizers

--hypervolume-optimizer-type [string, default = “adam”] Optimizer type for the hypervolume (choice: “adam”).

--pose-table-optimizer-type [string, default = “adam”] Optimizer for the pose table (choice: “adam”, “lbfgs”).

--conf-table-optimizer-type [string, default = “adam”] Optimizer for the conformation table (choice: “adam”).

--conf-encoder-optimizer-type [string, default = “adam”] Optimizer for the conformation encoder.

--lr [float, default = 1e-4] Learning rate of the hypervolume.

--lr-pose_table [float, default = 1e-3] Learning rate of the pose table.

--lr-conf-table [float, default = 1e-2] Learning rate fo the conformation table.

--lr-conf-encoder [float, default = 1e-4] Learning rate of the conformation encoder.

--wd [float, default = 0.0] Weight decay used by Adam optimizers.

Masking

--output-mask [string, default = “circ”] Type of output mask (choice: "circ", “frequency_marching”)

--add-one-frequency-every [int, default = 100,000] Frequency for adding new frequencies in the output mask, during HPS (in images).

--n-frequencies-per-epoch [int, default = 10] Number of frequencies to add in the output mask at each epoch, during SGD.

--max-freq [int, default = None] Highest frequency to use in the loss. Use all frequencies if None.

--window-radius-gt-real [float, default = 0.85] Radius of the circular mask applied on images in real space (maximum radius is 1).

--l-start-fm [int, default = 12] Starting size for the output mask when using frequency marching.

Conformations

--variational-het [bool, default = False] Flag activating the variational mode of conformation estimation.

--std-z-init [float, default = 0.1] Standard deviation of the initial conformations (i.i.d. with a centered Gaussian distribution).

--use-conf-encoder [bool, default = False] Flag for using an encoder to predict conformations.

--depth-cnn [int, default = 5] Depth of the encoder.

--channels-cnn [int, default = 32] Number of channels in the encoder.

--kernel-size-cnn [int, default = 3] Size of the kernels in the encoder.

--resolution-encoder [int, default = None] Resolution of images given to the encoder. Images are nto downsampled if None.

Hypervolume

--explicit-volume [bool, default = False] Flag for using an explicit volume (voxel array).

--layers [int, default = 3] Number of hidden layers in the hypervolume.

--dim [int, default = 256] Dimension of hidden layers in the hypervolume.

--pe-type [string, default = “gaussian”] Type of positional encoding for Fourier coordinates (choice: “gaussian”).

--pe-dim [int, default = 64] Number of frequencies used for positional encoding.

--feat-sigma [float, default = 0.5] Standard deviation of encoding frquencies.

--hypervolume-domain [string, default = “hartley”] Domain of the hypervolume (choice: “hartley”).

--pe-type-conf [string, default = None] Type of positional encoding for conformations (choice: None, “geom”).

Pre-training

--n-imgs-pretrain [int, default = 10,000] Number of images used for pre-training.

--pretrain-with-gt-poses [bool, default = False] Flag for using ground truth poses at pre-training time.

Pose Search

--l-start [int, default = 12] Number of frequencies used during the first pose search step.

--l-end [int, default = 32] Number of frequencies used during the last pose search step.

--niter [int, default = 4] Number of pose search iterations.

--t-extent [float, default = 20.0] Extent of the translation seach grid, in pixels.

--t-ngrid [int, default = 7] Number of point per dimension in the translation search grid.

--t-xshift [float, default = 0.0] X-axis shift of the translation search grid.

--t-yshift [float, default = 0.0] Y-axis shift of the translation search grid.

--no-trans-search-at-pose-search [bool, default = False] Flag for by-passing the translation search.

--nkeptposes [int, default = 8] Number of poses kept per image.

--base-healpy: [int, default = 2] Base healpy index.

Others

--color-palette [string, default = None] Type of palette used to visualize the predicted conformations (choice: None, “rainbow”, “linear”).

--no-amp Flag for disabling torch.cuda.amp Automatic Mixed Precision.

--test-installation [bool, default = False] Flag for testing the installation.

--seed [int, default = -1] Random seed. Randomly chosen if negative.

--verbose [bool, default = False] Flag for more verbose model logging.

--multigpu [bool, default = False] Flag for activating multi-GPU mode if more than one GPU is available.

--checkpoint [int, default = 5] Outputs from every nth epoch will saved, along with every pose search epoch.

--log-interval [int, default=10000] Logging interval in number of images.