viencoder/config.pbtxt

name: "viencoder"
# maximum batch size 
max_batch_size: 0 
platform: "ensemble"

#input to the model 
input [
{
  name: "TEXT"
  data_type: TYPE_STRING
  dims: [ -1 ] 
  # -1 means dynamic axis, aka this dimension may change 
}
]

#output of the model 
output {
  name: "output_0"
  data_type: TYPE_FP32
  dims: [-1, -1] 
  # two dimensional tensor, where 1st dimension: batch-size, 2nd dimension: #classes
}

#Type of scheduler to be used
ensemble_scheduling {
  step [
    {
      model_name: "viencoder.tokenizer"
      model_version: -1
      input_map {
      key: "TEXT"
      value: "TEXT"
    }
    output_map [
    {
      key: "input_ids"
      value: "input_ids"
    },
    {
      key: "attention_mask"
      value: "attention_mask"
    }
    ]
    },
    {
      model_name: "viencoder.model"
      model_version: -1
    input_map [
      {
        key: "input_ids"
        value: "input_ids"
      },
      {
        key: "attention_mask"
        value: "attention_mask"
      }
    ]
    output_map {
        key: "sentence_embedding"
        value: "output_0"
      }
    }
  ]
}