config.json

{
  "architecture": "ViT-Base + TrOCR decoder",
  "image_size": 384,
  "decoder_start_token_id": 0,
  "eos_token_id": 2,
  "pad_token_id": 1,
  "max_new_tokens": 256,
  "variants": {
    "int8": {
      "encoder": "models/int8/encoder.onnx",
      "decoder": "models/int8/decoder.onnx"
    },
    "fp16": {
      "encoder": "models/fp16/encoder.onnx",
      "decoder": "models/fp16/decoder.onnx"
    }
  }
}