config.json

{
  "absolute_positional_embedding_type": null,
  "architectures": [
    "TiteForPreTraining"
  ],
  "dropout_prob": 0.1,
  "hidden_act": "gelu_pytorch_tanh",
  "hidden_sizes": [
    768,
    768,
    768,
    1024,
    1024,
    1024,
    1280,
    1280,
    1280,
    1536,
    1536,
    1536
  ],
  "initializer_range": 0.02,
  "intermediate_sizes": [
    3072,
    3072,
    3072,
    4096,
    4096,
    4096,
    5120,
    5120,
    5120,
    6144,
    6144,
    6144
  ],
  "kernel_sizes": [
    null,
    null,
    null,
    2,
    2,
    2,
    2,
    2,
    2,
    2,
    2,
    2
  ],
  "layer_norm_eps": 1e-12,
  "max_position_embeddings": 512,
  "model_type": "tite",
  "norm_location": "post",
  "norm_type": "layer",
  "num_attention_heads": [
    12,
    12,
    12,
    16,
    16,
    16,
    20,
    20,
    20,
    24,
    24,
    24
  ],
  "num_hidden_layers": 12,
  "pad_token_id": 0,
  "pooling_implementation": "triton",
  "pooling_location": "intra",
  "positional_embedding_type": null,
  "relative_positional_embedding_type": "rotary",
  "rope_implementation": "eager",
  "rotary_interleaved": true,
  "strides": [
    null,
    null,
    null,
    2,
    2,
    2,
    2,
    2,
    2,
    2,
    2,
    2
  ],
  "torch_dtype": "float32",
  "transformers_version": "4.52.4",
  "vocab_size": 30522
}