v0.3.4

• Added --preserve-read-ids CLI flag to allow users to retain original read names from input data in the output BLOW5 or POD5 files
  • For BLOW5, original read IDs are written as plain strings (BLOW5 format permits this).
  • For POD5, original read IDs are deterministically converted to valid UUIDs using uuid.uuid5 (required by the POD5 spec), ensuring compatibility with downstream tools like Dorado.

What's Changed

• Fix/UUID reads by @denisbeslic in #23

Full Changelog: v0.3.3...v0.3.4

v0.3.3

Added

• Sequencing kit and flow cell info to .blow5 / .pod5 output (enables auto model basecalling, closes #20)
• min_read_len parameter to allow use of very short reference sequences

Changed

• Fixed --read-mode and -n: now uses the correct number of reads
• Updated preprocessing for latest Polars version
• Updated package dependencies

What's Changed

• v0.3.3 by @denisbeslic in #21

Full Changelog: v0.3.2...v0.3.3

v0.3.2

Added

• Multiple reads simulation: Added the ability to simulate multiple reads in read-mode, see #17. Corresponding README updates included.
• Added a new optional --bps parameter
• Expanded RNA support: Added RNA profiles RNA-004 and RNA-002 to the list of available simulation profiles. Extended preprocessing and training pipelines to handle RNA data. (Note: models are not available yet)

Changed

• Default noise parameter updated: The default value for noise-std was increased from 1.0 to 2.0, resulting in better performance on DNA datasets (see issues #17 and #18).
• Added detailed descriptions in the README covering training workflows and preprocessing steps for better usability.
• Introduced a warning when the average read length is shorter than the reference sequence.
• Seed parameter handling: When the seed parameter is not set, seq2squiggle now automatically selects a random seed to ensure variability in simulations (see #18).
• Library upgrades: Updated PyTorch and other dependencies to their latest versions for improved stability and performance.

What's Changed

• v0.3.2 by @denisbeslic in #19

Full Changelog: v0.3.1...v0.3.2

v0.3.1

v0.3.1 Release

Changed

• Documentation: Add bioconda installation guide
• Documentation: Update reference to publication

v0.3.0

v0.3.0 Release

Added

• Added latest R9.4.1 and R10.4.1 models
• Added Contributing Guide
• Advanced Parameters: Introduced a --show-advanced-options flag for experimental parameters in prediction mode.
• R9.4.1 Support: Added an R9.4.1 model and R9.4.1 profile for extended compatibility.
• Dwell Time Parameters: Added dwell-mean and dwell-std parameters for users preferring static distributions for dwell time generation.
• Model Download Simplification: Models can now be downloaded using the --profile argument, closes #7

Changed

• Noise Generation: Switched from NumPy to PyTorch for noise generation, ensuring consistent dependencies.
• Length Regulator: Fully transitioned to PyTorch, removing reliance on NumPy/Numba.
• Performance Boosts: Achieved small speedups by minimizing GPU-CPU data transfers and avoiding NumPy usage wherever possible.
• Preprocessing: Simplified preprocessing functions for improved maintainability.
• UUID Generation: Automatically generate a UUID4 in genome mode for unique identifiers and compability with dorado

v0.2.0

First PyPi release

Added

• function to automatically download the latest weights from GitHub
• __ version __ for version tracking
• version function to print the tool’s version
• Packaged the project with Poetry and prepared it for PyPI release

Changed

• Moved config.yaml to the src/ directory.
• Updated the README with the latest instructions and details
• Removed the wandb logger from the prediction process

Fixed

• Enabled prediction mode for CPU-only usage

Full Changelog: v0.1.0...v0.2.0

v0.1.0

v0.1.0 - 2024-08-13

Initial release