fastai is a deep learning library which provides users with high-level components that can quickly and easily provide state-of-the-art results in standard deep learning domains, and provides researchers with low-level components that can be mixed and matched to build new approaches.
It aims to provide both functionality without compromising ease of use, flexibility or performance. This is achieved through a carefully planned architecture, which expresses common patterns of many deep learning and data processing techniques. This allows users to leverage the flexibility of the PyTorch library, on which fastai has been build.
Artificial Intelligence engineer and Data Scientist who is passionate about science.
change batch tfms to have the correct dimensionality
A user found a bug in one of the course examples given by fastai. We were able to quickly fix this bug since the reporter also stated where the bug was located.
delete dev_nbs folder (see course-v3) #3273 (#3274)
Proposed was to delete this seemingly duplicate folder, since course-v3 covers this folder. Maintainability is harder when there are more than one location covering the same.
After interaction with the founder of the project, it was made clear that the fastai-library contained notebooks compatible with v2 of fastai, whereas the other was fastai v1 compatible.
I agree with the founder that it should be kept, however I think structure-wise the repository could be improved. I would propose to put it in a separate repository or rename the fastai folder.
In the documentation courses a link was pointing to a non-existing file with a wrong extension, so the extension was corrected
In the documentation about the use of git, an example was outdated. I found a new example, and updated the tutorial commands.
Fix dead link #3191
A dead link was found by another user. The fix of the user was invalid, so we proposed a quick fix.
implement quantile loss function
We have implemented a loss function that is not naturally supported through keras or any other machine learning library in general. The pinball loss function can be used to create quantile forecasts.