def get_dfs_from_hpt()

in sagemaker/source/visualization/model_visualisation_utils.py [0:0]

• 30 lines of code
• 4 McCabe index (conditional complexity)

def get_dfs_from_hpt(summaries, metrics):
    '''
    Helper function to get a list of dataframes from a HyperparameterTuningJobAnalytics summary
    Parameters:
    -----------
    summaries: {}
        Output of training_job_summaries() of the HyperparameterTuningJobAnalytics object
        
    metrics: [str]
        A list of names of the metrics (e.g., "test_auc")
        
    Returns:
    --------
    res: [(str, pd.DataFrame)]
        A list of dataframe where str is the jobname and pd.DataFrame is the correponding data.
    '''
    res = []
    for summary in summaries:
        job_name = summary["TrainingJobName"]
        job_df = pd.DataFrame()

        for m in metrics:
            df = TrainingJobAnalytics(job_name, [m], period=TIME_PERIOD).dataframe()
            df.rename(columns={'value':'{}'.format(m)}, inplace=True)
            if len(df) == 0:
                continue
            del df["metric_name"]
            timestamp = df["timestamp"]
            del df["timestamp"]
            # Ensure that there are at least 500 epochs
            job_df = pd.concat([job_df, df], 1)
            job_df["timestamp"] = timestamp/60
        res.append((job_name, job_df))
    return res