To wrap up, let’s use the model to test a prediction.
We are going to identify the class pair (predicted versus actual label) that the model struggles with the most, select one sample image from that pair, run a prediction and then show the image as ASCII art. That should give us a clue why the model struggles with that particular pair.
Let’s see if our AI agent can write all code in one go. Enter the following prompt:
...
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