** UPDATE: The videos for the course are now out!! Take a look at the fast.ai site for details: https://course.fast.ai/ **
We're back for round 2 of our ML workgroup and we will remain online for the duration of this course!!
Over the next few months we will work through the new version of the fast.ai course, "Practical Deep Learning For Coders v4, Part 1", this time using the latest version of the fast.ai api and the Fastbook.
Lesson 6 of the course (https://course.fast.ai/videos/?lesson=6) covers 3 notebooks, viz:
3. https://github.com/fastai/fastbook/blob/master/08_collab.ipynb (started)
We will tackle 05_pet_breeds.ipynb and 06_multicat, covering: how to get your data ready for training (looking at the datablocks in more detail and augmentation, etc), fitting your models using the learning rate finder, discriminative learning, cross entropy loss and fast.ai learners in more detail.
The format has changed - we will provide a very brief overview of the key contents of the notebook and walk through the questions in a 30-40 min session.
18:45 - Social chat, introduce yourself, questions, etc
18:59 - Overview: Maria & Perusha
19:00 - 05_pet_breeds and 06_multicat
19:40 - Team Projects:
- Les: ECG project
- Kalidass: GANs demo
Links for this lesson:
Course content is here: https://course.fast.ai/
Fastbook is here: https://github.com/fastai/fastbook
Fastbook notebooks without the text (just code blocks) are here: https://github.com/fastai/fastbook/tree/master/clean
How to get the most out of the course:
- Commit time every week - the videos/notebooks take a lot of time to digest and work through.
- Suggest starting with 2-3 hours for the video. The videos are dense so you want to stop it and make sure you understand as you go. Making notes in your copy of the fastbook may be useful for later review.
- Often 2-3 notebooks are covered in one video. Each notebook can take a lot of time to work through so allocate at least 1-2 hours per notebook
- If you are stuck ask questions on the slack channel or search the fast.ai forums for similar questions. The fast.ai community is vast so someone is bound to be able to help especially now with the general release of the course
- People are expected to have completed the material in their own time/pace before the event to be able to fully participate.
- We will space the online events 3-4 weeks apart to give people sufficient time to complete the assignments. It is likely that we will remain online until it is safe for everyone to re-emerge!