How to debug the explicability module of a model in the registry ?
0
Hello, I get an error in the explicability module of the model registry (see screenshot below). How can I debug this ? I would like to access logs to investigate this error.
2回答
- 新しい順
- 投票が多い順
- コメントが多い順
0
Customer needs to provide the S3uri for Clarify Explainability json in CreateModelPackage ModelMetrics
"ModelMetrics": {
"ModelQuality": {
"Statistics": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********/V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/modelqualitycheckstep/statistics.json"
},
"Constraints": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********//V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/modelqualitycheckstep/constraints.json"
}
},
"ModelDataQuality": {
"Statistics": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********//V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/dataqualitycheckstep/statistics.json"
},
"Constraints": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********//V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/dataqualitycheckstep/constraints.json"
}
},
"Bias": {
"Report": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********//V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/modelbiascheckstep/analysis.json"
},
"PreTrainingReport": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********/V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/databiascheckstep/analysis.json"
},
"PostTrainingReport": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********/V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/modelbiascheckstep/analysis.json"
}
},
"Explainability": {
"Report": {
"ContentType": "application/json",
"S3Uri": "s3://sagemaker-project-p-*********/V796221526-p-pwcowxgrrfdt/5qo6ezikmux3/modelexplainabilitycheckstep/analysis.json"
}
}
}
The Studio looks for the analysis.json(S3Uri) file in particular in the Explainability -> Report object and renders the report for the data in the analysis.json file.
回答済み 1年前
-1
If you are using XGBoost check this link: https://stackoverflow.com/questions/55621967/feature-importance-for-xgboost-in-sagemaker
回答済み 1年前
Thank for your answer but this Stackoverflow discussion is not about the same subject. I know how to get the importance of my model features from a notebook but in my case it is a custom model (catboost) that I save in the model registry while running a sagemaker pipeline. I can't find in the documentation the specification needed to display the feature importance when using AWS Clarify and I can't find the associated logs.