{
  "attention_probs_dropout_prob": 0.1,
  "config_tensor": {
    "attn": {
      "fc": {
        "amp": true,
        "factorization": "tt",
        "per_decomp_rank_ratio_limit": 0.75,
        "ranks": 300,
        "build_rank_parameters": false,
        "set_scale_factors": false,
        "shape": [
          16,
          8,
          8,
          8,
          8,
          16
        ],
        "tensorized": true
      },
      "sub_layers": {
        "amp": true,
        "factorization": "tt",
        "per_decomp_rank_ratio_limit": 0.75,
        "ranks": 300,
        "build_rank_parameters": false,
        "set_scale_factors": false,
        "shape": [
          16,
          8,
          8,
          8,
          8,
          16
        ],
        "tensorized": true
      }
    },
    "embedding": {
      "tensorized": false
    },
    "head": {
      "tensorized": false
    },
    "pff": [
      {
        "amp": true,
        "factorization": "tt",
        "per_decomp_rank_ratio_limit": 0.75,
        "ranks": 300,
        "build_rank_parameters": false,
        "set_scale_factors": false,
        "shape": [
          16,
          8,
          8,
          16,
          16,
          16
        ],
        "tensorized": true
      },
      {
        "amp": true,
        "factorization": "tt",
        "per_decomp_rank_ratio_limit": 0.75,
        "ranks": 300,
        "build_rank_parameters": false,
        "set_scale_factors": false,
        "shape": [
          16,
          16,
          16,
          8,
          8,
          16
        ],
        "tensorized": true
      }
    ],
    "pooler": {
      "tensorized": false
    }
  },
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.1,
  "hidden_size": 1024,
  "initializer_range": 0.02,
  "intermediate_size": 4096,
  "max_position_embeddings": 512,
  "num_attention_heads": 16,
  "num_hidden_layers": 24,
  "output_all_encoded_layers": false,
  "type_vocab_size": 2,
  "vocab_size": 30528
}