TypeError in FoodVision

[gauravreddy08]

Hey Dan, for some reason the notebooks stopped working and its giving an error when I run this cell

My guess is they changed the parameters for EfficientNet.

BTW, its not even working on the course notebook

[gauravreddy08]

Hey Daniel,

I figured it out, its probably a bug with the tensorflow 2.5.0 I downgraded the version to tensorflow 2.4.1 and it works fine now.

Weird that TensorFlow actually didn't look over this. I hope that you update the course materials, as new students who may come across this may face the same issue.

[gauravreddy08]

Notebook : Milestone Project 1 : FoodVision

And also Daniel does this issue come under bug with tensorflow. If so do I benefit anything by reporting it to Tensorflow?

[mrdbourke]

Okay looks like downgrading from 2.5.0 -> 2.4.1 fixes the issues.

My best guess is it will be fixed quite soon in 2.5.X

And also Daniel does this issue come under bug with tensorflow. If so do I benefit anything by reporting it to Tensorflow?

The best thing to do would be to post an issue (and explain where/how it happened) on the TensorFlow GitHub & talk with the developers there: https://github.com/tensorflow/tensorflow/issues

[gauravreddy08]

Yeah already did that 😄 tensorflow/tensorflow#49725

[ghost]

I was having this same issue yesterday. Thank you for figuring this out.
In addition, no matter how many times I tried I was not able to get a NVidia T4 to run the code, so I upgraded to Colaboratory Pro. Even then I could never get access to a T4 but was granted access to a Tesla V100-SXM2-16GB.

[mrdbourke]

@fredrider hmm, yeah I think having access to the right GPU is going to be a tough one (for those without Colab Pro).

I think the only reason I can access a T4 fairly regularly is that I'm in an Australian timezone and when I connect to Google Colab, more resources are available (due to other people around the world not being on).

But I decided to include the material on mixed precision because it's best practice.

Glad to hear you got it working though, if more people run into GPU issues, I'll add a note to the course.

[mrdbourke]

@ashikashikshafi08 has also raised a flag on the TensorFlow discussions forums: https://discuss.tensorflow.org/t/typeerror-input-y-of-sub-op-has-type-float16-that-does-not-match-type-float32-of-argument-x/1398

Adding the link above for future reference.