Wiki-Test2 / config.json
PhysiQuanty's picture
Duplicate from PhysiQuanty/Patent-Test-Radix-65536-AutoTokenizer_FineTune
efda231
raw
history blame contribute delete
603 Bytes
{
"model_type": "binaryllm",
"architectures": ["BinaryLLMForCausalLM"],
"auto_map": {
"AutoConfig": "configuration_binaryllm.BinaryLLMConfig",
"AutoModelForCausalLM": "modeling_binaryllm.BinaryLLMForCausalLM",
"AutoTokenizer": "tokenization_binaryllm.BinaryLLMTokenizer"
},
"vocab_size": 65538,
"bos_token_id": 65536,
"eos_token_id": 65537,
"pad_token_id": 65537,
"hidden_size": 512,
"num_hidden_layers": 4,
"num_attention_heads": 4,
"intermediate_size": 2048,
"max_position_embeddings": 2048,
"dropout": 0.1,
"activation": "gelu",
"torch_dtype": "float32"
}