Would it be helpful to have a wiki post to compile best-practices/rules-of-thumb/common-methodologies (with links)?
Knowing FastAI approach, prerequisite, and find your pathway is essential:
Fastbook – for whom?
FastAI Forum – which channel for whom, would it be helpful to have some kind of label for beginners, intermediate or have 10-grade level. What this strategy and other strategies can help the FastAI Forum user. Use them more effectively.
FastAI and NLP – specific topics.
Jeremy’s FastAI Coding Style:
Great place about to start with Best Practices:
Understand the structure of FastAI in nutshell snapshots:
30 Best practices suggested:
- Ask questions:
Probably, you can google or some good practice in stackoverflow or somewhere else for general ai practice, programming, etc. If you can think about something very specific for Fastai. Please add more here.
Figure out what is the uniqueness of Fastai and yourself. Good practice by asking questions:
a. How fastai is a good match for you?
b. What is your purpose of learning FastAI?
c. What is your learning rate -Lr?
d. What is your learning model, cf. ResNetX, architecture for what purpose and improvement/fine tuning?
e. What is your pattern and rhyme of your learning style?
f. What can be the loss rate of your learning?
g. How do you evaluate the accuracy and ways of validate what you learned?
h. How would you apply universal approximation theorem in your learning journey?
i. How would you deal with the data and info you get from Fastai?
j. What more questions can be helpful for you and others?
- Be inspired by FastAi that we can learn from about learning models and the process of developing AI fast or slow as our own learning journey, which can be extended much further beyond fastai. Thus, Thank you FastAI and the community.
Paraphrase of Jeremy’s Comment about the paper:
- Familiar stuff looks complex
- Important beyond basic, but hard to read
- Look for a blog about the paper
- Learn Greek alphabet(which is used in the paper) in Wikipedia or detexify.