v0.6.8

Fix the bug that the duplicate key shared by the last batch and the batch before the last

v0.5.3 bug fix: wds braceexpand style filename

Full Changelog: 0.5.2...0.5.3

v0.5.2 Add get_genome_size and random_coords functions

Add get_genome_size and random_coords functions, also the corresponding test functions

Full Changelog: 0.5.1...0.5.2

v0.5.1

add braceexpand option to dump_data_webdataset to simplify wds file list

v0.5.0: Utility functions for prepraring training dataset

Add several utility functions to help easily prepare the training dataset:

• filter_chromosomes
• make_random_shift
• make_flank
• chop_genome
• dna2OneHot
• rev_comp

0.4.0: function to transform existing webdataset file, param to create DALI compatible webdataset

• add a new function convert_data_webdataset to transform an existing webdataset file
• add DALI option to dump_data_webdataset to create NVIDIA DALI compatible webdataset for loading

Full Changelog: 0.3.0...0.4.0

0.3.0: add SeqChromDatasetByDataFrame function

New SeqChromDatasetByDataFrame function to load tensors given a region dataframe

What's Changed

• Sphinx document integration by @yztxwd in #1

New Contributors

• @yztxwd made their first contribution in #1

Full Changelog: 0.2.4...0.3.0

0.2.4: SeqchromDatasetByWds accepts device rank info, bug fix

1. Now SeqchromDatasetByWds can use device rank info to split files
2. Fix the bug occurred when there are more workers than # wds files
3. Allow empty Bigwig file list as input

Full Changelog: 0.2.3...0.2.4

Performance issue quick fix

Fix the performance issue due to each worker of webdataset loader uncompressed all files instead of uncompressing only the files assigned

Full Changelog: 0.2.2...0.2.3

Several bug fix: Duplicate webdataset sample index, default dataloader kwargs value

Fixes:

1. Add unique id to each sample to avoid duplicate sample name in webdataset
2. Change default dataloader kwargs to {}
3. Initialize Exception class before calling __init__()

Full Changelog: 0.2.1...0.2.2