Suggestion on Page411

[baifanhorst]

On page 411, several accuracies are added as class members in the MultiLayerPerceptron. However, when I tried to create an instance of this class, there occurred an error:

new() missing 1 required positional argument: 'task'

One possible solution is to specify several arguments for the accuracy function, as:

self.accuracy_train = torchmetrics.Accuracy(task="multiclass", num_classes=10)
self.accuracy_valid = torchmetrics.Accuracy(task="multiclass", num_classes=10)
self.accuracy_test = torchmetrics.Accuracy(task="multiclass", num_classes=10)

Namely, we should supply two arguments: task and num_classes. Also note that the names in these codes are slightly different from those in the book, so just change 'torchmetrics.Accuracy()' to 'torchmetrics.Accuracy(task="multiclass", num_classes=10)'.

[rasbt]

Good call, I should add a version check in the notebook to include the task and number of classes if a user is using a newer version of torchmetrics than specified in the book. Thanks for mentioning.

I also briefly saw the headlines of your other issues and will go through them one by one when time permits! Many thanks for submitting these!

[baifanhorst]

Thank you so much for your reply! I am just a beginner and a slow learner. I am following your book line by line to study machine learning.

…

On Mon, May 8, 2023 at 10:48 AM Sebastian Raschka ***@***.***> wrote: Good call, I should add a version check in the notebook to include the task and number of classes if a user is using a newer version of torchmetrics than specified in the book. Thanks for mentioning. I also briefly saw the headlines of your other issues and will go through them one by one when time permits! Many thanks for submitting these! — Reply to this email directly, view it on GitHub <#123 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3PUIM7R3PT7BIPN3KBB2NDXFEBTXANCNFSM6AAAAAAXZKTPRM> . You are receiving this because you authored the thread.Message ID: ***@***.*** com>

-- Sincerely, Fan. Fan Bai Florida State University, USA Department of Mathematics ***@***.*** ***@***.*** ***@***.***