what is the meaning of max_lr = slice(1e-6, 1e-4) in learn.fit_one_cycle(2, max_lr = slice(1e-6, 1e-4)). max_lr does it mean maximum leraning rate, why should we set maximum leraning rate as a slice(1e-6, 1e-4)?
Why we use the function learn.unfrezee() after learn.fit_one_cycle even though the model was not frozen?
Hi, I work for a film director who wants to work with a coder/programmer familiar with machine learning to use already existing code in a big music video in the future. Would any of you be interested in getting involved? Here are the references:
Please let me know, if you are interested in more details and to chat about the job please write me at song-i@bafic.co.uk
Thank you so much for providing this body of resources for making Deep Learning slightly more approachable in a very flexible way. As a Geographer well versed in GIS, I discovered Artificial Intelligence and Deep Learning out of necessity. Although well versed in Python and GIS, I am emphatically NOT well versed in high level math, but recognize the potential. This led me to discover Fast.AI, the Fast.AI MOOC, and these forums. Even after this first lesson, I am really enjoying digging around in this, and actually seeing results. While nothing to write home about, as a kayaker, I decided to see if I could train a model to recognize the difference between whitewater, recreational and sea kayak photos scraped from Google. After some input data cleanup, I was able to get the error rate under 0.01 - something I consider a win. Not a bad start!
Hi all,
Since I could not find a way to post a question or topic here, I put my question here.
When I run learn_critic.fit_one_cycle(6, 1e-3) on superres-gan.ipynb, I got an error which said RuntimeError: Expected object of scalar type Bool but got scalar type Byte for argument #2 'other'.
Sounds like training set went through without error but validation set caused the error.
Here’s my environments.
fastai version 1.0.57
pytorch 1.2.0
It’s too late, but I’ll answer just in case if someone else is curios about this.
fast.ai is not with Tensorflow but with Pytorch. You’ll be able to execute the lessons of fast ai in colab.research.google.com. But my advice is to go with Pytorch and/or Keras. Avoid Tensorflow - it’s tedious to learn and develop and is quite slow
See also this: https://www.youtube.com/watch?v=J6XcP4JOHmk
Thanks for the youtube link. I have been browsing through the forum to find my way around. I am new to learning this way and I was wondering if buying some textbooks might be a good idea. A friend suggested Artificial Intelligence: A Modern Approach. Do I really need it while doing this course? Is it worth the expenditure? I would rather conserve my resources.I would be happy with any advice that I could get.