% Generated by roxygen2: do not edit by hand

% Please edit documentation in R/generator_utils.R

\name{dataset_from_gen}

\alias{dataset_from_gen}

\title{Collect samples from generator and store in rds or pickle file.}

\usage{

dataset_from_gen(

  output_path,

  iterations = 10,

  train_type = "lm",

  output_format = "target_right",

  path_corpus,

  batch_size = 32,

  maxlen = 250,

  step = NULL,

  vocabulary = c("a", "c", "g", "t"),

  shuffle = FALSE,

  set_learning = NULL,

  seed = NULL,

  random_sampling = FALSE,

  store_format = "rds",

  file_name_start = "batch_",

  masked_lm = NULL,

  ...

)

}

\arguments{

\item{output_path}{Output directory. Output files will be named \code{output_path} + \code{file_name_start} + x + ".rds" or ".pickle", where x is an index (from 1 to

\code{iterations}) and file ending depends on \code{store_format} argument.}

\item{iterations}{Number of batches (output files) to create.}

\item{train_type}{Either \code{"lm"}, \code{"lm_rds"}, \code{"masked_lm"} for language model; \code{"label_header"}, \code{"label_folder"}, \code{"label_csv"}, \code{"label_rds"} for classification or \code{"dummy_gen"}.

\itemize{

\item Language model is trained to predict character(s) in a sequence. \cr

\item \code{"label_header"}/\code{"label_folder"}/\code{"label_csv"} are trained to predict a corresponding class given a sequence as input.

\item If \code{"label_header"}, class will be read from fasta headers.

\item If \code{"label_folder"}, class will be read from folder, i.e. all files in one folder must belong to the same class.

\item If \code{"label_csv"}, targets are read from a csv file. This file should have one column named "file". The targets then correspond to entries in that row (except "file"

column). Example: if we are currently working with a file called "a.fasta" and corresponding label is "label_1", there should be a row in our csv file\tabular{lll}{

   file \tab label_1 \tab label_2 \cr

   "a.fasta" \tab 1 \tab 0 \cr

}

\item If \code{"label_rds"}, generator will iterate over set of .rds files containing each a list of input and target tensors. Not implemented for model

with multiple inputs.

\item If \code{"lm_rds"}, generator will iterate over set of .rds files and will split tensor according to \code{target_len} argument

(targets are last \code{target_len} nucleotides of each sequence).

\item  If \code{"dummy_gen"}, generator creates random data once and repeatedly feeds these to model.

\item  If \code{"masked_lm"}, generator maskes some parts of the input. See \code{masked_lm} argument for details.

}}

\item{output_format}{Determines shape of output tensor for language model.

Either \code{"target_right"}, \code{"target_middle_lstm"}, \code{"target_middle_cnn"} or \code{"wavenet"}.

Assume a sequence \code{"AACCGTA"}. Output correspond as follows

\itemize{

\item \verb{"target_right": X = "AACCGT", Y = "A"}

\item \verb{"target_middle_lstm": X = (X_1 = "AAC", X_2 = "ATG"), Y = "C"} (note reversed order of X_2)

\item \verb{"target_middle_cnn": X = "AACGTA", Y = "C"}

\item \verb{"wavenet": X = "AACCGT", Y = "ACCGTA"}

}}

\item{path_corpus}{Input directory where fasta files are located or path to single file ending with fasta or fastq

(as specified in format argument). Can also be a list of directories and/or files.}

\item{batch_size}{Number of samples in one batch.}

\item{maxlen}{Length of predictor sequence.}

\item{step}{How often to take a sample.}

\item{vocabulary}{Vector of allowed characters. Characters outside vocabulary get encoded as specified in \code{ambiguous_nuc}.}

\item{shuffle}{Whether to shuffle samples within each batch.}

\item{set_learning}{When you want to assign one label to set of samples. Only implemented for \code{train_type = "label_folder"}.

Input is a list with the following parameters

\itemize{

\item \code{samples_per_target}: how many samples to use for one target.

\item \code{maxlen}: length of one sample.

\item \code{reshape_mode}: \verb{"time_dist", "multi_input"} or \code{"concat"}.

\itemize{

\item

If \code{reshape_mode} is \code{"multi_input"}, generator will produce \code{samples_per_target} separate inputs, each of length \code{maxlen} (model should have

\code{samples_per_target} input layers).

\item If reshape_mode is \code{"time_dist"}, generator will produce a 4D input array. The dimensions correspond to

\verb{(batch_size, samples_per_target, maxlen, length(vocabulary))}.

\item If \code{reshape_mode} is \code{"concat"}, generator will concatenate \code{samples_per_target} sequences

of length \code{maxlen} to one long sequence.

}

\item If \code{reshape_mode} is \code{"concat"}, there is an additional \code{buffer_len}

argument. If \code{buffer_len} is an integer, the subsequences are interspaced with \code{buffer_len} rows. The input length is

(\code{maxlen} \eqn{*} \code{samples_per_target}) + \code{buffer_len} \eqn{*} (\code{samples_per_target} - 1).

}}

\item{seed}{Sets seed for \code{set.seed} function for reproducible results.}

\item{random_sampling}{Whether samples should be taken from random positions when using \code{max_samples} argument. If \code{FALSE} random

samples are taken from a consecutive subsequence.}

\item{store_format}{Either "rds" or "pickle".}

\item{file_name_start}{Start of output file names.}

\item{masked_lm}{If not \code{NULL}, input and target are equal except some parts of the input are masked or random.

Must be list with the following arguments:

\itemize{

\item \code{mask_rate}: Rate of input to mask (rate of input to replace with mask token).

\item \code{random_rate}: Rate of input to set to random token.

\item \code{identity_rate}: Rate of input where sample weights are applied but input and output are identical.

\item \code{include_sw}: Whether to include sample weights.

\item \code{block_len} (optional): Masked/random/identity regions appear in blocks of size \code{block_len}.

}}

\item{...}{further generator options. See \code{\link{get_generator}}.}

}

\value{

None. Function writes data to files and does not return a value.

}

\description{

Repeatedly generate samples with data generator and store output. Creates a separate rds or pickle file in \code{output_path} for each

batch.

}

\examples{

\dontshow{if (reticulate::py_module_available("tensorflow")) (if (getRversion() >= "3.4") withAutoprint else force)(\{ # examplesIf}

# create dummy fasta files

temp_dir <- tempfile()

dir.create(temp_dir)

create_dummy_data(file_path = temp_dir,

                  num_files = 3,

                  seq_length = 8, 

                  num_seq = 2)

# extract samples

out_dir <- tempfile()

dir.create(out_dir)

dataset_from_gen(output_path = out_dir,

                 iterations = 10,

                 train_type = "lm",

                 output_format = "target_right",

                 path_corpus = temp_dir, 

                 batch_size = 32,

                 maxlen = 5,

                 step = 1,

                 file_name_start = "batch_")

list.files(out_dir)

\dontshow{\}) # examplesIf}

}