|
{ |
|
"_from_model_config": true, |
|
"_name_or_path": "openbmb/cpm-bee-10b", |
|
"architectures": [ |
|
"CpmBeeForCausalLM" |
|
], |
|
"auto_map": { |
|
"AutoConfig": "openbmb/cpm-bee-10b--configuration_cpmbee.CpmBeeConfig", |
|
"AutoModel": "openbmb/cpm-bee-10b--modeling_cpmbee.CpmBeeForCausalLM", |
|
"AutoModelForCausalLM": "openbmb/cpm-bee-10b--modeling_cpmbee.CpmBeeForCausalLM" |
|
}, |
|
"dim_ff": 10240, |
|
"dim_head": 128, |
|
"distance_scale": 16, |
|
"dropout_p": 0.0, |
|
"eps": 1e-06, |
|
"half": true, |
|
"hidden_size": 4096, |
|
"init_std": 1.0, |
|
"mask_modules": [ |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
], |
|
[ |
|
false, |
|
false |
|
] |
|
], |
|
"model_type": "cpmbee", |
|
"num_attention_heads": 32, |
|
"num_hidden_layers": 48, |
|
"position_bias_max_distance": 2048, |
|
"position_bias_num_buckets": 256, |
|
"position_bias_num_segment_buckets": 256, |
|
"torch_dtype": "float16", |
|
"transformers_version": "4.30.2", |
|
"use_cache": true, |
|
"vocab_size": 86583 |
|
} |
|
|