{
  "architectures": [
    "ReformerModelWithLMHead"
  ],
  "attention_probs_dropout_prob": 0.3,
  "chunk_length": 7,
  "chunk_size_feed_forward": 0,
  "chunk_size_lm_head": 0,
  "d_ff": 128,
  "d_model": 128,
  "feed_forward_size": 128,
  "ff_chunk_size": 0,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.3,
  "hidden_size": 128,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "layer_norm_eps": 1e-12,
  "max_position_embeddings": 512,
  "model_type": "bert",
  "num_attention_chunks": 1,
  "num_attention_heads": 2,
  "num_buckets": 2,
  "num_chunks": 0,
  "num_chunks_after": 0,
  "num_chunks_before": 1,
  "num_hashes": 4,
  "num_hidden_layers": 3,
  "output_past": true,
  "seed": 0,
  "sinusoidal_pos_embds": true,
  "type_vocab_size": 2,
  "vocab_size": 200
}