fuel
fuel copied to clipboard
mila-iqia
A data pipeline framework for machine learning
@yaoli was interested in using a divide-and-conquer approach to preprocessing, as is used in in @dwf's ImageNet PR (https://github.com/bartvm/fuel/pull/68). With that code, I think it should be relatively easy to...
TextFile dataset docstring mention that dictionary could be a path to the dictionary pickled file, but never inside the code pickle loading exists.
This is something that could be part of the handshake between the server and the client.
Please add support for UTF8 files in TextFile (i.e use codecs.open and allow the user to pass in an encoding). If you like, I could submit this as a patch,...
Leads to weird errors when trying to do attribute things prior to calling super().
Seems like integers or lists of integers, but I couldn't find this written down anywhere. Slice object should eventually be supported but perhaps not by default due to the silent...
Following offline discussion with @lamblin and @vdumoulin, we agreed that the most important kind of type checking to perform in the data processing pipeline is probably the semantics of the...
Here's a minimal example reproducing the bug: ``` python import numpy from fuel.datasets import IndexableDataset from fuel.streams import DataStream from fuel.schemes import SequentialScheme from fuel.transformers import Mapping def run(bug=True): constructor...
Fuel currently depends on both PyTables and h5py, using them for different datasets. Ideally I think we should at least make PyTables an optional dependency.