Isaak-Carter's picture
removed the modeling codes
912ef0b verified
{
"architectures": [
"MiniCPMForCausalLM"
],
"bos_token_id": 1,
"dim_model_base": 256,
"eos_token_id": 2,
"hidden_act": "silu",
"hidden_size": 2304,
"initializer_range": 0.1,
"intermediate_size": 5760,
"max_position_embeddings": 4096,
"model_type": "minicpm",
"num_attention_heads": 36,
"num_hidden_layers": 40,
"num_key_value_heads": 36,
"quantization": {
"group_size": 64,
"bits": 4
},
"rms_norm_eps": 1e-05,
"rope_scaling": null,
"scale_depth": 1.4,
"scale_emb": 12,
"torch_dtype": "bfloat16",
"transformers_version": "4.36.0",
"use_cache": true,
"vocab_size": 122753
}