{
  "model_name": "HyformerForDownstreamPrediction",
  "bias": false,
  "attention_dropout": 0.0,
  "feed_forward_dropout": 0.0,
  "embedding_dim": 512,
  "embedding_hidden_dim": 2048,
  "layer_norm_eps": 1e-05,
  "max_seq_len": 128,
  "num_heads": 8,
  "num_layers": 12,
  "vocab_size": 596,
  "num_physchem_tasks": 200,
  "pooler_dropout": 0.0,
  "pooler_hidden_dim": 512
}