encoderfile 0.6.2-rc.2

Distribute and run transformer encoders with a single file.
Documentation
syntax = "proto3";

package encoderfile.embedding;

import "proto/token.proto";
import "proto/metadata.proto";

service EmbeddingInference {
  rpc Predict(EmbeddingRequest) returns (EmbeddingResponse);
  rpc GetModelMetadata(encoderfile.metadata.GetModelMetadataRequest) returns (encoderfile.metadata.GetModelMetadataResponse);
}

message EmbeddingRequest {
  repeated string inputs = 1;
  map<string, string> metadata = 3;
}

message EmbeddingResponse {
  // len(embeddings) == len(inputs)
  repeated TokenEmbeddingSequence results = 1;
  string model_id = 2;
  map<string, string> metadata = 3;
}

message TokenEmbeddingSequence {
  repeated TokenEmbedding embeddings = 1;
}

message TokenEmbedding {
  repeated float embedding = 1;
  encoderfile.token.TokenInfo token_info = 2;
}