pretrained/MiniCPM-2B-dpo-bf16/model_config.json

{
    "attn_pdrop": 0.0,
    "bos_token_id": 1,
    "eos_token_id": 2,
    "pad_token_id": null,
    "hidden_act": "silu",
    "hidden_size": 2304,
    "initializer_range": 0.1,
    "intermediate_size": 5760,
    "max_position_embeddings": 2048,
    "num_attention_heads": 36,
    "num_hidden_layers": 40,
    "num_key_value_heads": 36,
    "pretraining_tp": 1,
    "rms_norm_eps": 1e-05,
    "rope_theta": 10000.0,
    "tie_word_embeddings": true,
    "use_cache": true,
    "vocab_size": 122753,
    "scale_emb": 12,
    "dim_model_base": 256,
    "scale_depth": 1.4
}