phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-23 05:22:26 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Update API docs

Browse source
This commit is contained in:
Sai Soundararaj 2025-07-02 11:34:34 -07:00
parent 4d0d2d685f
commit 8f96b61c43
26 changed files with 1397 additions and 32 deletions
Download patch file Download diff file Expand all files Collapse all files

33
llama_stack/apis/synthetic_data_generation/synthetic_data_generation.py
View file

@ -14,7 +14,15 @@ from llama_stack.schema_utils import json_schema_type, webmethod
class FilteringFunction(Enum):
"""The type of filtering function."""
"""The type of filtering function.
:cvar none: No filtering applied, accept all generated synthetic data
:cvar random: Random sampling of generated data points
:cvar top_k: Keep only the top-k highest scoring synthetic data samples
:cvar top_p: Nucleus-style filtering, keep samples exceeding cumulative score threshold
:cvar top_k_top_p: Combined top-k and top-p filtering strategy
:cvar sigmoid: Apply sigmoid function for probability-based filtering
"""
none = "none"
random = "random"
@ -26,7 +34,12 @@ class FilteringFunction(Enum):
@json_schema_type
class SyntheticDataGenerationRequest(BaseModel):
"""Request to generate synthetic data. A small batch of prompts and a filtering function"""
"""Request to generate synthetic data. A small batch of prompts and a filtering function
:param dialogs: List of conversation messages to use as input for synthetic data generation
:param filtering_function: Type of filtering to apply to generated synthetic data samples
:param model: (Optional) The identifier of the model to use. The model must be registered with Llama Stack and available via the /models endpoint
"""
dialogs: list[Message]
filtering_function: FilteringFunction = FilteringFunction.none
@ -35,7 +48,11 @@ class SyntheticDataGenerationRequest(BaseModel):
@json_schema_type
class SyntheticDataGenerationResponse(BaseModel):
"""Response from the synthetic data generation. Batch of (prompt, response, score) tuples that pass the threshold."""
"""Response from the synthetic data generation. Batch of (prompt, response, score) tuples that pass the threshold.
:param synthetic_data: List of generated synthetic data samples that passed the filtering criteria
:param statistics: (Optional) Statistical information about the generation process and filtering results
"""
synthetic_data: list[dict[str, Any]]
statistics: dict[str, Any] | None = None
@ -48,4 +65,12 @@ class SyntheticDataGeneration(Protocol):
dialogs: list[Message],
filtering_function: FilteringFunction = FilteringFunction.none,
model: str | None = None,
) -> SyntheticDataGenerationResponse: ...
) -> SyntheticDataGenerationResponse:
"""Generate synthetic data based on input dialogs and apply filtering.
:param dialogs: List of conversation messages to use as input for synthetic data generation
:param filtering_function: Type of filtering to apply to generated synthetic data samples
:param model: (Optional) The identifier of the model to use. The model must be registered with Llama Stack and available via the /models endpoint
:returns: Response containing filtered synthetic data samples and optional statistics
"""
...