kz930 opened a new issue, #8058: URL: https://github.com/apache/texera/issues/8058
### What happened? The operator hands the two configured columns straight to scikit-learn's metric functions: ```python result['Accuracy'][0] = round(accuracy_score(y_true, y_pred), 4) prediction = metrics_func[metric](y_true, y_pred, average=None, labels=[label]) ``` When either column has an empty cell, that value reaches the metric as a `None`, and scikit-learn refuses the pair: `ValueError: Classification metrics can't handle a mix of unknown and binary targets`. The run ends, and the message names neither the column nor a way out. Nothing in the configuration can exclude the row. The regression branch is the same shape, calling `mean_squared_error` and `r2_score` on the same two columns. I expected it to score the rows it can and leave the unusable ones out, the way the visualization operators drop missing values before plotting and the Hugging Face operators keep the row with an empty result. ### How to reproduce? 1. Build a table with an actual-value column and a predicted-value column where one row has a blank in either. 2. Connect Machine Learning Scorer, pointing its two attributes at those columns and selecting any classification metric. 3. Run. The run ends with the ValueError below. With every cell filled it completes normally. ### Version/Branch 1.3.0-incubating-SNAPSHOT (main) ### Relevant log output ```shell ValueError: Classification metrics can't handle a mix of unknown and binary targets ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
