{
  "_name_or_path": "/mnt/efs/people/dejiaoz/experiments/next_models/codesage_v2_pub/cs-small-clv7-fv1-3e05-4k18k-mlmdobf-para-homo-t003-mrl/combined_bf16_unwrapped.pt",
  "activation_function": "gelu_new",
  "architectures": [
    "CodeSageModel"
  ],
  "attention_dropout_prob": 0.1,
  "auto_map": {
    "AutoConfig": "config_codesage.CodeSageConfig",
    "AutoModel": "modeling_codesage.CodeSageModel",
    "AutoModelForMaskedLM": "modeling_codesage.CodeSageForMaskedLM",
    "AutoModelForSequenceClassification": "modeling_codesage.CodeSageForSequenceClassification"
  },
  "bos_token_id": 100257,
  "embedding_dropout_prob": 0.1,
  "eos_token_id": 100257,
  "hidden_size": 1024,
  "initializer_range": 0.02,
  "intermediate_size": 4096,
  "layer_norm_epsilon": 1e-05,
  "max_position_embeddings": 2048,
  "model_type": "codesage",
  "num_attention_heads": 8,
  "num_hidden_layers": 6,
  "pad_token_id": 100317,
  "position_embedding_type": "absolute",
  "residual_dropout_prob": 0.1,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.26.1",
  "vocab_size": 49154
}