Galileo API 422: Field logits must be an array of floats matching number of classes

A 422 means the syntax is fine but the data shape is wrong. For a classification model, the logits array for every input must have exactly one float per class. The docs example: with 3 classes and label 2, logits of [[1.2, 0.8]] fails because it only has 2 values; [[1.2, 0.8, -0.3]] passes. So count your classes and make sure every row of logits has exactly that many floats. The same idea applies to tokens and labels: the model type you declared determines the required input structure, and every logged row has to match it.

Context: POSTing model outputs to the Galileo API returns 422 unprocessable entity: "Field logits must be an array of floats matching number of classes." My payload is valid JSON. What shape does Galileo expect?