EvalsAPI Images Inputs: Fix "grading_results" construction: it's .sample, not ["sample"]

[gabor-openai]

Summary

EvalsAPI Images Inputs notebook

1. Fix: "grading_results" construction: it's .sample, not ["sample"]
2. Improve: Add "score" and "passed" to the dataframe for easier interpretability

Motivation

The cell I've changed previously errored out with:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[18], line 11
      3 if run.status == "completed" or run.status == "failed": # check if the run is finished
      4     output_items = list(client.evals.runs.output_items.list(
      5         run_id=run.id, eval_id=eval_object.id
      6     ))
      7     df = pd.DataFrame({
      8             "prompt": [item.datasource_item["prompt"]for item in output_items],
      9             "reference": [item.datasource_item["reference"] for item in output_items],
     10             "model_response": [item.sample.output[0].content for item in output_items],
---> 11             "grading_results": [item.results[0]["sample"]["output"][0]["content"]
     12                                 for item in output_items]
     13         })
     14     display(df)
     15     break

TypeError: 'Result' object is not subscriptable