All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
0.6.2 - 2024-14-09
- bump
tch-rs
to 0.17 - update
itertools
to 0.13 - use new style modules
- implement
ExactSizeIterator
when possible (thanks @AzHicham) - change the
Fetch
trait to use&self
instead of&mut self
(thanks @AzHicham)
- add typos to the ci
- add cargo deny to the ci
0.6.1 - 2023-13-08
- fix parallelalisation,
install
wasn't sufficient to run things in a multithreaded way.
- avoid resetting the thread pools when the number of threads are the same.
0.6.0 - 2023-22-05
- add support for closure as collate function.
- add basic parallelalisation (feat @AzHicham)
0.5.4 - 2023-29-04
- online doc with feature tch are broken
0.5.3 - 2023-27-04
- online doc didn't show all features
0.5.2 - 2023-27-04
- basic benchmark with criterion.
- improve doc by showing features.
0.5.1 - 2023-25-04
- primitive type for
TorchCollate
are now collated intch::Tensor
instead ofndarray
. - clippy pedantic lints
0.5.0 - 2023-18-04
- refactor example organisation
- rename features
torch
totch
. - make tch loading on the GPU opt-in
- bunp tch to 0.11.0
0.4.0 - 2023-17-03
- Add tch-rs integration to run on the GPU.
0.3.1 - 2022-13-11
- Incorrect bound for iterable dataloader builder, which doesn't allowed provide a custom collate function.
0.3.0 - 2022-01-11
- more strict compiler warnings
Dataloader
for iterable datasets. A dalaloader working for any type that implementsIntoIterator
- shuffling support for iterable dataset
- project folder reorganization
- default collate now support reference
0.2.1 - 2022-27-09
- Fix link in the doc
- Add missing documentation field in package metadata
0.2.0 - 2022-27-09
- Indexable dataset
DefaultCollate
functionDataloader
andDataloaderBuilder
- Sequential and random
Sampler