clip/model_configs/RBT3-chinese.json

{
    "vocab_size": 21128,
    "text_attention_probs_dropout_prob": 0.1, 
    "text_hidden_act": "gelu", 
    "text_hidden_dropout_prob": 0.1, 
    "text_hidden_size": 768,
    "text_initializer_range": 0.02, 
    "text_intermediate_size": 3072, 
    "text_max_position_embeddings": 512, 
    "text_num_attention_heads": 12, 
    "text_num_hidden_layers": 3,
    "text_type_vocab_size": 2  
}