ccdv's picture
add mask_first_token
f160ab4
{
"_name_or_path": "ccdv/lsg-distilcamembert-base-4096",
"adaptive": true,
"architectures": [
"LSGCamembertForMaskedLM"
],
"attention_probs_dropout_prob": 0.1,
"auto_map": {
"AutoConfig": "modeling_lsg_camembert.LSGCamembertConfig",
"AutoModel": "modeling_lsg_camembert.LSGCamembertModel",
"AutoModelForCausalLM": "modeling_lsg_camembert.LSGCamembertForCausalLM",
"AutoModelForMaskedLM": "modeling_lsg_camembert.LSGCamembertForMaskedLM",
"AutoModelForMultipleChoice": "modeling_lsg_camembert.LSGCamembertForMultipleChoice",
"AutoModelForQuestionAnswering": "modeling_lsg_camembert.LSGCamembertForQuestionAnswering",
"AutoModelForSequenceClassification": "modeling_lsg_camembert.LSGCamembertForSequenceClassification",
"AutoModelForTokenClassification": "modeling_lsg_camembert.LSGCamembertForTokenClassification"
},
"base_model_prefix": "lsg",
"block_size": 128,
"bos_token_id": 0,
"classifier_dropout": null,
"eos_token_id": 2,
"gradient_checkpointing": false,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.1,
"hidden_size": 768,
"initializer_range": 0.02,
"intermediate_size": 3072,
"layer_norm_eps": 1e-05,
"lsh_num_pre_rounds": 1,
"mask_first_token": false,
"max_position_embeddings": 4098,
"model_type": "camembert",
"num_attention_heads": 12,
"num_global_tokens": 1,
"num_hidden_layers": 6,
"pad_token_id": 1,
"pool_with_global": true,
"position_embedding_type": "absolute",
"sparse_block_size": 128,
"sparsity_factor": 2,
"sparsity_type": "norm",
"torch_dtype": "float32",
"transformers_version": "4.19.2",
"type_vocab_size": 1,
"use_cache": true,
"vocab_size": 32005
}