{ "model_type": "custom", "architecture": "CharSeq2Seq", "task": "text2text-generation", "language": "non" }