Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add failure handling for set processors in ML inference pipelines #144654

Merged
merged 5 commits into from
Nov 16, 2022
Merged

Add failure handling for set processors in ML inference pipelines #144654

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
819a3f6
Add failure handling for set processors
brianmcgue Nov 4, 2022
8e71a24
Merge branch 'main' into bmcgue/on_failure_scenarios
brianmcgue Nov 14, 2022
84518ea
Use bracket style getter in painless script
brianmcgue Nov 15, 2022
e5eb975
Replace set processor on_failure with null checks
brianmcgue Nov 15, 2022
9d3f044
Merge branch 'main' into bmcgue/on_failure_scenarios
brianmcgue Nov 15, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
105 changes: 101 additions & 4 deletions x-pack/plugins/enterprise_search/common/ml_inference_pipeline/index.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,11 @@
* 2.0.
*/

import { IngestSetProcessor, MlTrainedModelConfig } from '@elastic/elasticsearch/lib/api/types';
import {
IngestRemoveProcessor,
IngestSetProcessor,
MlTrainedModelConfig,
} from '@elastic/elasticsearch/lib/api/types';
import { BUILT_IN_MODEL_TAG } from '@kbn/ml-plugin/common/constants/data_frame_analytics';
import { SUPPORTED_PYTORCH_TASKS } from '@kbn/ml-plugin/common/constants/trained_models';

Expand All @@ -18,6 +22,7 @@ import {
getSetProcessorForInferenceType,
SUPPORTED_PYTORCH_TASKS as LOCAL_SUPPORTED_PYTORCH_TASKS,
parseMlInferenceParametersFromPipeline,
getRemoveProcessorForInferenceType,
} from '.';

const mockModel: MlTrainedModelConfig = {
Expand Down Expand Up @@ -63,6 +68,38 @@ describe('getMlModelTypesForModelConfig lib function', () => {
});
});

describe('getRemoveProcessorForInferenceType lib function', () => {
const destinationField = 'dest';

it('should return expected value for TEXT_CLASSIFICATION', () => {
const inferenceType = SUPPORTED_PYTORCH_TASKS.TEXT_CLASSIFICATION;

const expected: IngestRemoveProcessor = {
field: destinationField,
ignore_missing: true,
};

expect(getRemoveProcessorForInferenceType(destinationField, inferenceType)).toEqual(expected);
});

it('should return expected value for TEXT_EMBEDDING', () => {
const inferenceType = SUPPORTED_PYTORCH_TASKS.TEXT_EMBEDDING;

const expected: IngestRemoveProcessor = {
field: destinationField,
ignore_missing: true,
};

expect(getRemoveProcessorForInferenceType(destinationField, inferenceType)).toEqual(expected);
});

it('should return undefined for unknown inferenceType', () => {
const inferenceType = 'wrongInferenceType';

expect(getRemoveProcessorForInferenceType(destinationField, inferenceType)).toBeUndefined();
});
});

describe('getSetProcessorForInferenceType lib function', () => {
const destinationField = 'dest';

Expand All @@ -79,10 +116,28 @@ describe('getSetProcessorForInferenceType lib function', () => {
"Copy the predicted_value to 'dest' if the prediction_probability is greater than 0.5",
field: destinationField,
if: 'ml.inference.dest.prediction_probability > 0.5',
on_failure: [
brianmcgue marked this conversation as resolved.
Show resolved Hide resolved
{
append: {
field: '_source._ingest.set_errors',
ignore_failure: true,
value: [
{
message:
"Processor 'set' in pipeline 'my-pipeline' failed with message '{{ _ingest.on_failure_message }}'",
pipeline: 'my-pipeline',
timestamp: '{{{ _ingest.timestamp }}}',
},
],
},
},
],
value: undefined,
};

expect(getSetProcessorForInferenceType(destinationField, inferenceType)).toEqual(expected);
expect(getSetProcessorForInferenceType(destinationField, inferenceType, 'my-pipeline')).toEqual(
expected
);
});

it('should return expected value for TEXT_EMBEDDING', () => {
Expand All @@ -92,16 +147,36 @@ describe('getSetProcessorForInferenceType lib function', () => {
copy_from: 'ml.inference.dest.predicted_value',
description: "Copy the predicted_value to 'dest'",
field: destinationField,
on_failure: [
{
append: {
field: '_source._ingest.set_errors',
ignore_failure: true,
value: [
{
message:
"Processor 'set' in pipeline 'my-pipeline' failed with message '{{ _ingest.on_failure_message }}'",
pipeline: 'my-pipeline',
timestamp: '{{{ _ingest.timestamp }}}',
},
],
},
},
],
value: undefined,
};

expect(getSetProcessorForInferenceType(destinationField, inferenceType)).toEqual(expected);
expect(getSetProcessorForInferenceType(destinationField, inferenceType, 'my-pipeline')).toEqual(
expected
);
});

it('should return undefined for unknown inferenceType', () => {
const inferenceType = 'wrongInferenceType';

expect(getSetProcessorForInferenceType(destinationField, inferenceType)).toBeUndefined();
expect(
getSetProcessorForInferenceType(destinationField, inferenceType, 'my-pipeline')
).toBeUndefined();
});
});

Expand Down Expand Up @@ -185,13 +260,35 @@ describe('generateMlInferencePipelineBody lib function', () => {
expect.objectContaining({
description: expect.any(String),
processors: expect.arrayContaining([
expect.objectContaining({
remove: {
field: 'my-destination-field',
ignore_missing: true,
},
}),
expect.objectContaining({
set: {
copy_from: 'ml.inference.my-destination-field.predicted_value',
description:
"Copy the predicted_value to 'my-destination-field' if the prediction_probability is greater than 0.5",
field: 'my-destination-field',
if: 'ml.inference.my-destination-field.prediction_probability > 0.5',
on_failure: [
{
append: {
field: '_source._ingest.set_errors',
ignore_failure: true,
value: [
{
message:
"Processor 'set' in pipeline 'my-pipeline' failed with message '{{ _ingest.on_failure_message }}'",
pipeline: 'my-pipeline',
timestamp: '{{{ _ingest.timestamp }}}',
},
],
},
},
],
},
}),
]),
Expand Down
41 changes: 39 additions & 2 deletions x-pack/plugins/enterprise_search/common/ml_inference_pipeline/index.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,8 @@

import {
IngestPipeline,
IngestProcessorContainer,
IngestRemoveProcessor,
IngestSetProcessor,
MlTrainedModelConfig,
} from '@elastic/elasticsearch/lib/api/types';
Expand Down Expand Up @@ -53,7 +55,8 @@ export const generateMlInferencePipelineBody = ({
model.input?.field_names?.length > 0 ? model.input.field_names[0] : 'MODEL_INPUT_FIELD';

const inferenceType = Object.keys(model.inference_config)[0];
const set = getSetProcessorForInferenceType(destinationField, inferenceType);
const remove = getRemoveProcessorForInferenceType(destinationField, inferenceType);
const set = getSetProcessorForInferenceType(destinationField, inferenceType, pipelineName);

return {
description: description ?? '',
Expand All @@ -64,6 +67,7 @@ export const generateMlInferencePipelineBody = ({
ignore_missing: true,
},
},
...(remove ? [{ remove }] : []),
{
inference: {
field_map: {
Expand Down Expand Up @@ -108,31 +112,64 @@ export const generateMlInferencePipelineBody = ({

export const getSetProcessorForInferenceType = (
destinationField: string,
inferenceType: string
inferenceType: string,
pipelineName: string
): IngestSetProcessor | undefined => {
let set: IngestSetProcessor | undefined;
const prefixedDestinationField = `ml.inference.${destinationField}`;
const onFailure: IngestProcessorContainer[] = [
{
append: {
field: '_source._ingest.set_errors',
ignore_failure: true,
value: [
{
message: `Processor 'set' in pipeline '${pipelineName}' failed with message '{{ _ingest.on_failure_message }}'`,
pipeline: pipelineName,
timestamp: '{{{ _ingest.timestamp }}}',
},
],
},
},
];

if (inferenceType === SUPPORTED_PYTORCH_TASKS.TEXT_CLASSIFICATION) {
set = {
copy_from: `${prefixedDestinationField}.predicted_value`,
description: `Copy the predicted_value to '${destinationField}' if the prediction_probability is greater than 0.5`,
field: destinationField,
if: `${prefixedDestinationField}.prediction_probability > 0.5`,
on_failure: onFailure,
value: undefined,
};
} else if (inferenceType === SUPPORTED_PYTORCH_TASKS.TEXT_EMBEDDING) {
set = {
copy_from: `${prefixedDestinationField}.predicted_value`,
description: `Copy the predicted_value to '${destinationField}'`,
field: destinationField,
on_failure: onFailure,
value: undefined,
};
}

return set;
};

export const getRemoveProcessorForInferenceType = (
destinationField: string,
inferenceType: string
): IngestRemoveProcessor | undefined => {
if (
inferenceType === SUPPORTED_PYTORCH_TASKS.TEXT_CLASSIFICATION ||
inferenceType === SUPPORTED_PYTORCH_TASKS.TEXT_EMBEDDING
) {
return {
field: destinationField,
ignore_missing: true,
};
}
};

/**
* Parses model types list from the given configuration of a trained machine learning model
* @param trainedModel configuration for a trained machine learning model
Expand Down