This commit is contained in:
Xi Yan 2025-03-12 00:13:27 -07:00
parent 817331e76e
commit 0abedd070c
3 changed files with 63 additions and 62 deletions

View file

@ -2518,7 +2518,7 @@
"tags": [ "tags": [
"Datasets" "Datasets"
], ],
"description": "Register a new dataset through a file or", "description": "Register a new dataset.",
"parameters": [], "parameters": [],
"requestBody": { "requestBody": {
"content": { "content": {
@ -6838,24 +6838,24 @@
], ],
"title": "Benchmark" "title": "Benchmark"
}, },
"DataReference": { "DataSource": {
"oneOf": [ "oneOf": [
{ {
"$ref": "#/components/schemas/URIDataReference" "$ref": "#/components/schemas/URIDataSource"
}, },
{ {
"$ref": "#/components/schemas/HuggingfaceDataReference" "$ref": "#/components/schemas/HuggingfaceDataSource"
}, },
{ {
"$ref": "#/components/schemas/RowsDataReference" "$ref": "#/components/schemas/RowsDataSource"
} }
], ],
"discriminator": { "discriminator": {
"propertyName": "type", "propertyName": "type",
"mapping": { "mapping": {
"uri": "#/components/schemas/URIDataReference", "uri": "#/components/schemas/URIDataSource",
"huggingface": "#/components/schemas/HuggingfaceDataReference", "huggingface": "#/components/schemas/HuggingfaceDataSource",
"rows": "#/components/schemas/RowsDataReference" "rows": "#/components/schemas/RowsDataSource"
} }
} }
}, },
@ -6879,13 +6879,13 @@
"schema": { "schema": {
"type": "string", "type": "string",
"enum": [ "enum": [
"jsonl_messages" "messages"
], ],
"title": "Schema", "title": "Schema",
"description": "Schema of the dataset. Each type has a different column format." "description": "Schema of the dataset. Each type has a different column format."
}, },
"data_reference": { "data_source": {
"$ref": "#/components/schemas/DataReference" "$ref": "#/components/schemas/DataSource"
}, },
"metadata": { "metadata": {
"type": "object", "type": "object",
@ -6920,12 +6920,12 @@
"provider_id", "provider_id",
"type", "type",
"schema", "schema",
"data_reference", "data_source",
"metadata" "metadata"
], ],
"title": "Dataset" "title": "Dataset"
}, },
"HuggingfaceDataReference": { "HuggingfaceDataSource": {
"type": "object", "type": "object",
"properties": { "properties": {
"type": { "type": {
@ -6968,9 +6968,9 @@
"dataset_path", "dataset_path",
"params" "params"
], ],
"title": "HuggingfaceDataReference" "title": "HuggingfaceDataSource"
}, },
"RowsDataReference": { "RowsDataSource": {
"type": "object", "type": "object",
"properties": { "properties": {
"type": { "type": {
@ -7012,9 +7012,9 @@
"type", "type",
"rows" "rows"
], ],
"title": "RowsDataReference" "title": "RowsDataSource"
}, },
"URIDataReference": { "URIDataSource": {
"type": "object", "type": "object",
"properties": { "properties": {
"type": { "type": {
@ -7031,7 +7031,7 @@
"type", "type",
"uri" "uri"
], ],
"title": "URIDataReference" "title": "URIDataSource"
}, },
"Model": { "Model": {
"type": "object", "type": "object",
@ -9383,13 +9383,13 @@
"schema": { "schema": {
"type": "string", "type": "string",
"enum": [ "enum": [
"jsonl_messages" "messages"
], ],
"description": "The schema format of the dataset. One of - jsonl_messages: The dataset is a JSONL file with messages in column format" "description": "The schema format of the dataset. One of - jsonl_messages: The dataset is a JSONL file with messages in column format"
}, },
"data_reference": { "data_source": {
"$ref": "#/components/schemas/DataReference", "$ref": "#/components/schemas/DataSource",
"description": "The data reference of the dataset. Examples: - { \"type\": \"uri\", \"uri\": \"https://mywebsite.com/mydata.jsonl\" } - { \"type\": \"uri\", \"uri\": \"lsfs://mydata.jsonl\" } - { \"type\": \"huggingface\", \"dataset_path\": \"tatsu-lab/alpaca\", \"params\": { \"split\": \"train\" } } - { \"type\": \"rows\", \"rows\": [{\"message\": \"Hello, world!\"}] }" "description": "The data source of the dataset. Examples: - { \"type\": \"uri\", \"uri\": \"https://mywebsite.com/mydata.jsonl\" } - { \"type\": \"uri\", \"uri\": \"lsfs://mydata.jsonl\" } - { \"type\": \"huggingface\", \"dataset_path\": \"tatsu-lab/alpaca\", \"params\": { \"split\": \"train\" } } - { \"type\": \"rows\", \"rows\": [ { \"messages\": [ {\"role\": \"user\", \"content\": \"Hello, world!\"}, {\"role\": \"assistant\", \"content\": \"Hello, world!\"}, ] } ] }"
}, },
"metadata": { "metadata": {
"type": "object", "type": "object",
@ -9425,7 +9425,7 @@
"additionalProperties": false, "additionalProperties": false,
"required": [ "required": [
"schema", "schema",
"data_reference" "data_source"
], ],
"title": "RegisterDatasetRequest" "title": "RegisterDatasetRequest"
}, },

View file

@ -1698,7 +1698,7 @@ paths:
$ref: '#/components/responses/DefaultError' $ref: '#/components/responses/DefaultError'
tags: tags:
- Datasets - Datasets
description: Register a new dataset through a file or description: Register a new dataset.
parameters: [] parameters: []
requestBody: requestBody:
content: content:
@ -4731,17 +4731,17 @@ components:
- scoring_functions - scoring_functions
- metadata - metadata
title: Benchmark title: Benchmark
DataReference: DataSource:
oneOf: oneOf:
- $ref: '#/components/schemas/URIDataReference' - $ref: '#/components/schemas/URIDataSource'
- $ref: '#/components/schemas/HuggingfaceDataReference' - $ref: '#/components/schemas/HuggingfaceDataSource'
- $ref: '#/components/schemas/RowsDataReference' - $ref: '#/components/schemas/RowsDataSource'
discriminator: discriminator:
propertyName: type propertyName: type
mapping: mapping:
uri: '#/components/schemas/URIDataReference' uri: '#/components/schemas/URIDataSource'
huggingface: '#/components/schemas/HuggingfaceDataReference' huggingface: '#/components/schemas/HuggingfaceDataSource'
rows: '#/components/schemas/RowsDataReference' rows: '#/components/schemas/RowsDataSource'
Dataset: Dataset:
type: object type: object
properties: properties:
@ -4758,12 +4758,12 @@ components:
schema: schema:
type: string type: string
enum: enum:
- jsonl_messages - messages
title: Schema title: Schema
description: >- description: >-
Schema of the dataset. Each type has a different column format. Schema of the dataset. Each type has a different column format.
data_reference: data_source:
$ref: '#/components/schemas/DataReference' $ref: '#/components/schemas/DataSource'
metadata: metadata:
type: object type: object
additionalProperties: additionalProperties:
@ -4781,10 +4781,10 @@ components:
- provider_id - provider_id
- type - type
- schema - schema
- data_reference - data_source
- metadata - metadata
title: Dataset title: Dataset
HuggingfaceDataReference: HuggingfaceDataSource:
type: object type: object
properties: properties:
type: type:
@ -4808,8 +4808,8 @@ components:
- type - type
- dataset_path - dataset_path
- params - params
title: HuggingfaceDataReference title: HuggingfaceDataSource
RowsDataReference: RowsDataSource:
type: object type: object
properties: properties:
type: type:
@ -4832,8 +4832,8 @@ components:
required: required:
- type - type
- rows - rows
title: RowsDataReference title: RowsDataSource
URIDataReference: URIDataSource:
type: object type: object
properties: properties:
type: type:
@ -4846,7 +4846,7 @@ components:
required: required:
- type - type
- uri - uri
title: URIDataReference title: URIDataSource
Model: Model:
type: object type: object
properties: properties:
@ -6346,18 +6346,19 @@ components:
schema: schema:
type: string type: string
enum: enum:
- jsonl_messages - messages
description: >- description: >-
The schema format of the dataset. One of - jsonl_messages: The dataset The schema format of the dataset. One of - jsonl_messages: The dataset
is a JSONL file with messages in column format is a JSONL file with messages in column format
data_reference: data_source:
$ref: '#/components/schemas/DataReference' $ref: '#/components/schemas/DataSource'
description: >- description: >-
The data reference of the dataset. Examples: - { "type": "uri", "uri": The data source of the dataset. Examples: - { "type": "uri", "uri": "https://mywebsite.com/mydata.jsonl"
"https://mywebsite.com/mydata.jsonl" } - { "type": "uri", "uri": "lsfs://mydata.jsonl" } - { "type": "uri", "uri": "lsfs://mydata.jsonl" } - { "type": "huggingface",
} - { "type": "huggingface", "dataset_path": "tatsu-lab/alpaca", "params": "dataset_path": "tatsu-lab/alpaca", "params": { "split": "train" } } -
{ "split": "train" } } - { "type": "rows", "rows": [{"message": "Hello, { "type": "rows", "rows": [ { "messages": [ {"role": "user", "content":
world!"}] } "Hello, world!"}, {"role": "assistant", "content": "Hello, world!"}, ]
} ] }
metadata: metadata:
type: object type: object
additionalProperties: additionalProperties:
@ -6377,7 +6378,7 @@ components:
additionalProperties: false additionalProperties: false
required: required:
- schema - schema
- data_reference - data_source
title: RegisterDatasetRequest title: RegisterDatasetRequest
RegisterModelRequest: RegisterModelRequest:
type: object type: object

View file

@ -16,7 +16,7 @@ from llama_stack.schema_utils import json_schema_type, register_schema, webmetho
class Schema(Enum): class Schema(Enum):
""" """
Schema of the dataset. Each type has a different column format. Schema of the dataset. Each type has a different column format.
:cvar jsonl_messages: The dataset is a JSONL file with messages. Examples: :cvar messages: The dataset contains messages used for post-training. Examples:
{ {
"messages": [ "messages": [
{"role": "user", "content": "Hello, world!"}, {"role": "user", "content": "Hello, world!"},
@ -25,7 +25,7 @@ class Schema(Enum):
} }
""" """
jsonl_messages = "jsonl_messages" messages = "messages"
# TODO: add more schemas here # TODO: add more schemas here
@ -36,36 +36,36 @@ class DatasetType(Enum):
@json_schema_type @json_schema_type
class URIDataReference(BaseModel): class URIDataSource(BaseModel):
type: Literal["uri"] = "uri" type: Literal["uri"] = "uri"
uri: str uri: str
@json_schema_type @json_schema_type
class HuggingfaceDataReference(BaseModel): class HuggingfaceDataSource(BaseModel):
type: Literal["huggingface"] = "huggingface" type: Literal["huggingface"] = "huggingface"
dataset_path: str dataset_path: str
params: Dict[str, Any] params: Dict[str, Any]
@json_schema_type @json_schema_type
class RowsDataReference(BaseModel): class RowsDataSource(BaseModel):
type: Literal["rows"] = "rows" type: Literal["rows"] = "rows"
rows: List[Dict[str, Any]] rows: List[Dict[str, Any]]
DataReference = register_schema( DataSource = register_schema(
Annotated[ Annotated[
Union[URIDataReference, HuggingfaceDataReference, RowsDataReference], Union[URIDataSource, HuggingfaceDataSource, RowsDataSource],
Field(discriminator="type"), Field(discriminator="type"),
], ],
name="DataReference", name="DataSource",
) )
class CommonDatasetFields(BaseModel): class CommonDatasetFields(BaseModel):
schema: Schema schema: Schema
data_reference: DataReference data_source: DataSource
metadata: Dict[str, Any] = Field( metadata: Dict[str, Any] = Field(
default_factory=dict, default_factory=dict,
description="Any additional metadata for this dataset", description="Any additional metadata for this dataset",
@ -100,16 +100,16 @@ class Datasets(Protocol):
async def register_dataset( async def register_dataset(
self, self,
schema: Schema, schema: Schema,
data_reference: DataReference, data_source: DataSource,
metadata: Optional[Dict[str, Any]] = None, metadata: Optional[Dict[str, Any]] = None,
dataset_id: Optional[str] = None, dataset_id: Optional[str] = None,
) -> Dataset: ) -> Dataset:
""" """
Register a new dataset through a file or Register a new dataset.
:param schema: The schema format of the dataset. One of :param schema: The schema format of the dataset. One of
- jsonl_messages: The dataset is a JSONL file with messages in column format - jsonl_messages: The dataset is a JSONL file with messages in column format
:param data_reference: The data reference of the dataset. Examples: :param data_source: The data source of the dataset. Examples:
- { - {
"type": "uri", "type": "uri",
"uri": "https://mywebsite.com/mydata.jsonl" "uri": "https://mywebsite.com/mydata.jsonl"