Other problem in chapter 4

📚 Official Books Machine Learning by Tutorials
1

Hi,

When I execute this code in jupyter notebook:

model = tc.image_classifier.create(train_data, target=“label”,
model=“VisionFeaturePrint_Screen”,
verbose=True,
max_iterations=50)

I have the error:

ToolkitError: VisionFeaturePrint_Screen is not a valid option for model. Expected one of: resnet-50, squeezenet_v1.1

It seems despite in the book it supports three it seems it only supports 2.

Doing conda list it says I have 5.0 version:

packages in environment at /anaconda3/envs/turienv:

Name Version Build Channel

turicreate 5.0

1 Like
2

I have macOS 10.13.6, but I can’t update to macOS 10.14 because my macbook pro is of 2011. May be the cause that I can’t use this model. Is there any way to solve it?

3

@audrey @hollance Can you please help with this when you get a chance? Thank you - much appreciated! :]

4

hi Felix: unfortunately, VisionFeaturePrint_Screen requires macOS 10.14

https://apple.github.io/turicreate/docs/api/generated/turicreate.image_classifier.create.html#turicreate.image_classifier.create

you can use resnet or squeezenet but the model will be larger

btw don’t use the full data set, or it will take forever to run; it took hours on my older MBP, and yours is even older.

1 Like