phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

update EvaluationTask

Browse source
This commit is contained in:
Xi Yan 2025-03-18 19:28:34 -07:00
parent 5e817cd56a
commit f107e3229b
3 changed files with 56 additions and 195 deletions
Download patch file Download diff file Expand all files Collapse all files

90
docs/_static/llama-stack-spec.html vendored
View file

@ -2035,7 +2035,7 @@
]
}
},
"/v1/evaluation/grading": {
"/v1/evaluation/grade": {
"post": {
"responses": {
"200": {
@ -8523,32 +8523,14 @@
],
"title": "VectorDB"
},
"BenchmarkEvaluationTask": {
"EvaluationTask": {
"type": "object",
"properties": {
"type": {
"type": "string",
"const": "benchmark",
"default": "benchmark"
},
"benchmark_id": {
"type": "string"
}
},
"additionalProperties": false,
"required": [
"type",
"benchmark_id"
],
"title": "BenchmarkEvaluationTask"
},
"DataEvaluationTask": {
"type": "object",
"properties": {
"type": {
"type": "string",
"const": "data",
"default": "data"
},
"dataset_id": {
"type": "string"
},
"data_source": {
"$ref": "#/components/schemas/DataSource"
@ -8561,66 +8543,14 @@
}
},
"additionalProperties": false,
"required": [
"type",
"data_source",
"grader_ids"
],
"title": "DataEvaluationTask"
},
"DatasetEvaluationTask": {
"type": "object",
"properties": {
"type": {
"type": "string",
"const": "dataset",
"default": "dataset"
},
"dataset_id": {
"type": "string"
},
"grader_ids": {
"type": "array",
"items": {
"type": "string"
}
}
},
"additionalProperties": false,
"required": [
"type",
"dataset_id",
"grader_ids"
],
"title": "DatasetEvaluationTask"
},
"EvaluationTask": {
"oneOf": [
{
"$ref": "#/components/schemas/BenchmarkEvaluationTask"
},
{
"$ref": "#/components/schemas/DatasetEvaluationTask"
},
{
"$ref": "#/components/schemas/DataEvaluationTask"
}
],
"discriminator": {
"propertyName": "type",
"mapping": {
"benchmark": "#/components/schemas/BenchmarkEvaluationTask",
"dataset": "#/components/schemas/DatasetEvaluationTask",
"data": "#/components/schemas/DataEvaluationTask"
}
}
"title": "EvaluationTask"
},
"GradeRequest": {
"type": "object",
"properties": {
"task": {
"$ref": "#/components/schemas/EvaluationTask",
"description": "The task to evaluate. One of: - BenchmarkEvaluationTask: Run evaluation task against a benchmark_id - DatasetEvaluationTask: Run evaluation task against a dataset_id and a list of grader_ids - DataEvaluationTask: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
"description": "The task to evaluate. To specify a task, one of the following must be provided: - `benchmark_id`: Run evaluation task against a benchmark_id - `dataset_id` and `grader_ids`: Run evaluation task against a dataset_id and a list of grader_ids - `data_source` and `grader_ids`: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
}
},
"additionalProperties": false,
@ -8706,7 +8636,7 @@
"properties": {
"task": {
"$ref": "#/components/schemas/EvaluationTask",
"description": "The task to evaluate. One of: - BenchmarkEvaluationTask: Run evaluation task against a benchmark_id - DatasetEvaluationTask: Run evaluation task against a dataset_id and a list of grader_ids - DataEvaluationTask: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
"description": "The task to evaluate. To specify a task, one of the following must be provided: - `benchmark_id`: Run evaluation task against a benchmark_id - `dataset_id` and `grader_ids`: Run evaluation task against a dataset_id and a list of grader_ids - `data_source` and `grader_ids`: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
}
},
"additionalProperties": false,
@ -10737,7 +10667,7 @@
"properties": {
"task": {
"$ref": "#/components/schemas/EvaluationTask",
"description": "The task to evaluate. One of: - BenchmarkEvaluationTask: Run evaluation task against a benchmark_id - DatasetEvaluationTask: Run evaluation task against a dataset_id and a list of grader_ids - DataEvaluationTask: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
"description": "The task to evaluate. To specify a task, one of the following must be provided: - `benchmark_id`: Run evaluation task against a benchmark_id - `dataset_id` and `grader_ids`: Run evaluation task against a dataset_id and a list of grader_ids - `data_source` and `grader_ids`: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
},
"candidate": {
"$ref": "#/components/schemas/EvaluationCandidate",
@ -10839,7 +10769,7 @@
"properties": {
"task": {
"$ref": "#/components/schemas/EvaluationTask",
"description": "The task to evaluate. One of: - BenchmarkEvaluationTask: Run evaluation task against a benchmark_id - DatasetEvaluationTask: Run evaluation task against a dataset_id and a list of grader_ids - DataEvaluationTask: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
"description": "The task to evaluate. To specify a task, one of the following must be provided: - `benchmark_id`: Run evaluation task against a benchmark_id - `dataset_id` and `grader_ids`: Run evaluation task against a dataset_id and a list of grader_ids - `data_source` and `grader_ids`: Run evaluation task against a data source (e.g. rows, uri, etc.) and a list of grader_ids"
},
"candidate": {
"$ref": "#/components/schemas/EvaluationCandidate",