name: "postprocess"
backend: "python"
max_batch_size: 16

input [
  {
    name: "TOKEN_EMBEDS_POST"
    data_type: TYPE_FP32
    dims: [128, 384]
    
  },
   {
    name: "ATTENTION_POST"
    data_type: TYPE_INT32
    dims: [128]
  }
]
output [
  {
    name: "SENT_EMBED"
    data_type: TYPE_FP32
    dims: [ 384 ]
  }
]

instance_group [{ kind: KIND_GPU }]

parameters: {
  key: "EXECUTION_ENV_PATH",
  value: {string_value: "$$TRITON_MODEL_DIRECTORY/processing_env.tar.gz"}
}