private static LabelSuggestion Predict <T>(
T issueOrPr,
PredictionEngine <T, GitHubIssuePrediction> predEngine,
ILogger logger)
where T : IssueModel
{
if (predEngine == null)
{
throw new InvalidOperationException("expected prediction engine loaded.");
}
GitHubIssuePrediction prediction = predEngine.Predict(issueOrPr);
VBuffer <ReadOnlyMemory <char> > slotNames = default;
predEngine.OutputSchema[nameof(GitHubIssuePrediction.Score)].GetSlotNames(ref slotNames);
float[] probabilities = prediction.Score;
var labelPredictions = MikLabelerPredictor.GetBestThreePredictions(probabilities, slotNames);
float maxProbability = probabilities.Max();
logger.LogInformation($"# {maxProbability} {prediction.Area} for #{issueOrPr.Number} {issueOrPr.Title}");
return(new LabelSuggestion
{
LabelScores = labelPredictions,
});
}
public static async Task <string> PredictAsync(GitHubIssue issue, ILogger logger, double threshold)
{
PredictionModel <GitHubIssue, GitHubIssuePrediction> model = await PredictionModel.ReadAsync <GitHubIssue, GitHubIssuePrediction>(ModelPath);
GitHubIssuePrediction prediction = model.Predict(issue);
float[] probabilities = prediction.Probabilities;
float maxProbability = probabilities.Max();
logger.LogInformation($"# {maxProbability.ToString()} {prediction.Area} for #{issue.ID} {issue.Title}");
return(maxProbability > threshold ? prediction.Area : null);
}
public static async Task <string> PredictAsync(GitHubIssue issue, ILogger logger)
{
PredictionModel <GitHubIssue, GitHubIssuePrediction> model = await PredictionModel.ReadAsync <GitHubIssue, GitHubIssuePrediction>(ModelPath);
GitHubIssuePrediction prediction = model.Predict(issue);
float[] probabilities = prediction.Probabilities;
float maxProbability = probabilities.Max();
logger.LogInformation($"Label {prediction.Area} for {issue.ID} is predicted with confidence {maxProbability.ToString()}");
return(maxProbability > 0.8 ? prediction.Area : null);
}
public static string Predict(GitHubIssue issue, ILogger logger, double threshold)
{
if (predEngine == null)
{
MLContext mlContext = new MLContext();
ITransformer mlModel = mlContext.Model.Load(ModelPath, out DataViewSchema inputSchema);
predEngine = mlContext.Model.CreatePredictionEngine <GitHubIssue, GitHubIssuePrediction>(mlModel);
}
GitHubIssuePrediction prediction = predEngine.Predict(issue);
float[] probabilities = prediction.Score;
float maxProbability = probabilities.Max();
logger.LogInformation($"# {maxProbability.ToString()} {prediction.Area} for #{issue.Number} {issue.Title}");
return(maxProbability > threshold ? prediction.Area : null);
}
public static string Predict <T>(T issueOrPr, ref PredictionEngine <T, GitHubIssuePrediction> predEngine, ILogger logger, double threshold)
where T : IssueModel
{
if (predEngine == null)
{
MLContext mlContext = new MLContext();
ITransformer mlModel = mlContext.Model.Load(PrModelPath, out DataViewSchema inputSchema);
predEngine = mlContext.Model.CreatePredictionEngine <T, GitHubIssuePrediction>(mlModel);
}
GitHubIssuePrediction prediction = predEngine.Predict(issueOrPr);
float[] probabilities = prediction.Score;
float maxProbability = probabilities.Max();
string typeToPredict = issueOrPr is IssueModel ? "issue" : "PR";
logger.LogInformation($"# {maxProbability} {prediction.Area} for {typeToPredict} #{issueOrPr.Number} {issueOrPr.Title}");
return(maxProbability > threshold ? prediction.Area : null);
}
