Lumilio Photos

English

简体中文

English

简体中文

Appearance

server ​

gRPC server for ML prediction services.

This module initializes and serves a gRPC endpoint that delegates prediction requests to registered model services, such as CLIP.

ModelRegistry Objects ​

pml/server.py
python
class ModelRegistry()

Manage registration and lookup of ML model services.

Attributes:

  • services - Mapping of service names to service instances.
  • start_time - Timestamp when the registry was created.

register_service ​

pml/server.py
python
def register_service(name: str, service: Any) -> None

Add a model service to the registry.

Arguments:

  • name - Identifier for the service.
  • service - Instance providing the service interface.

get_service ​

pml/server.py
python
def get_service(name: str) -> Optional[Any]

Retrieve a registered service by name.

Arguments:

  • name - Identifier of the service to fetch.

Returns:

The service instance if found; otherwise None.

list_services ​

pml/server.py
python
def list_services() -> list

List all registered service names.

Returns:

A list of service identifiers.

get_uptime ​

pml/server.py
python
def get_uptime() -> int

Calculate uptime since registry initialization.

Returns:

Uptime in seconds.

PredictionServiceServicer Objects ​

pml/server.py
python
class PredictionServiceServicer(ml_service_pb2_grpc.PredictionServiceServicer)

gRPC servicer implementing prediction and health-check endpoints.

Delegates incoming calls to specific model services registered in the ModelRegistry.

ProcessImageForCLIP ​

pml/server.py
python
def ProcessImageForCLIP(request, context)

Handle image processing requests for CLIP.

Arguments:

  • request - Protobuf request containing raw image bytes.
  • context - gRPC context for status and metadata.

Returns:

ImageProcessResponse with processed image features.

GetTextEmbeddingForCLIP ​

pml/server.py
python
def GetTextEmbeddingForCLIP(request, context)

Handle text embedding requests for CLIP.

Arguments:

  • request - Protobuf request containing text input.
  • context - gRPC context for status and metadata.

Returns:

TextEmbeddingResponse with embedding vector.

Predict ​

pml/server.py
python
def Predict(request, context)

Generic prediction endpoint supporting multiple models.

Dispatches based on request.model_name. Supports CLIP for both image and text inputs.

Arguments:

  • request - PredictRequest containing model name and input data.
  • context - gRPC context for status and metadata.

Returns:

PredictResponse with float values and metadata.

BatchPredict ​

pml/server.py
python
def BatchPredict(request, context)

Process a batch of prediction requests.

Arguments:

  • request - BatchPredictRequest with multiple PredictRequests.
  • context - gRPC context for status and metadata.

Returns:

BatchPredictResponse summarizing successes and failures.

HealthCheck ​

pml/server.py
python
def HealthCheck(request, context)

Health check endpoint for individual or all services.

Arguments:

  • request - HealthCheckRequest specifying a service name or 'all'.
  • context - gRPC context for status and metadata.

Returns:

HealthCheckResponse with status, uptime, and message.

serve ​

pml/server.py
python
def serve(port: int = 50051, max_workers: int = 10) -> None

Start the gRPC server for ML prediction.

Arguments:

  • port - TCP port number to listen on.
  • max_workers - Maximum number of thread pool workers.

Copyright © 2025 EdwinZhanCN