U 4Af@sndZddlmZddlmZddlmZddlmZddl m Z e e Z Gdd d eZGd d d eZd S) zI-BERT configuration) OrderedDict)Mapping)PretrainedConfig) OnnxConfig)loggingcs&eZdZdZdZdfdd ZZS) IBertConfiga This is the configuration class to store the configuration of a [`IBertModel`]. It is used to instantiate a I-BERT model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the IBERT [kssteven/ibert-roberta-base](https://huggingface.co/kssteven/ibert-roberta-base) architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: vocab_size (`int`, *optional*, defaults to 30522): Vocabulary size of the I-BERT model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`IBertModel`] hidden_size (`int`, *optional*, defaults to 768): Dimensionality of the encoder layers and the pooler layer. num_hidden_layers (`int`, *optional*, defaults to 12): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 12): Number of attention heads for each attention layer in the Transformer encoder. intermediate_size (`int`, *optional*, defaults to 3072): Dimensionality of the "intermediate" (often named feed-forward) layer in the Transformer encoder. hidden_act (`str` or `Callable`, *optional*, defaults to `"gelu"`): The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, `"relu"`, `"silu"` and `"gelu_new"` are supported. hidden_dropout_prob (`float`, *optional*, defaults to 0.1): The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1): The dropout ratio for the attention probabilities. max_position_embeddings (`int`, *optional*, defaults to 512): The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048). type_vocab_size (`int`, *optional*, defaults to 2): The vocabulary size of the `token_type_ids` passed when calling [`IBertModel`] initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. layer_norm_eps (`float`, *optional*, defaults to 1e-12): The epsilon used by the layer normalization layers. position_embedding_type (`str`, *optional*, defaults to `"absolute"`): Type of position embedding. Choose one of `"absolute"`, `"relative_key"`, `"relative_key_query"`. For positional embeddings use `"absolute"`. For more information on `"relative_key"`, please refer to [Self-Attention with Relative Position Representations (Shaw et al.)](https://arxiv.org/abs/1803.02155). For more information on `"relative_key_query"`, please refer to *Method 4* in [Improve Transformer Models with Better Relative Position Embeddings (Huang et al.)](https://arxiv.org/abs/2009.13658). quant_mode (`bool`, *optional*, defaults to `False`): Whether to quantize the model or not. force_dequant (`str`, *optional*, defaults to `"none"`): Force dequantize specific nonlinear layer. Dequatized layers are then executed with full precision. `"none"`, `"gelu"`, `"softmax"`, `"layernorm"` and `"nonlinear"` are supported. As deafult, it is set as `"none"`, which does not dequantize any layers. Please specify `"gelu"`, `"softmax"`, or `"layernorm"` to dequantize GELU, Softmax, or LayerNorm, respectively. `"nonlinear"` will dequantize all nonlinear layers, i.e., GELU, Softmax, and LayerNorm. Zibert:w gelu皙?{Gz?-q=rabsoluteFnonec sxtjf| ||d|||_||_||_||_||_||_||_||_ | |_ | |_ | |_ | |_ ||_||_||_dS)N) pad_token_id bos_token_id eos_token_id)super__init__ vocab_size hidden_sizenum_hidden_layersnum_attention_heads hidden_actintermediate_sizehidden_dropout_probattention_probs_dropout_probmax_position_embeddingstype_vocab_sizeinitializer_rangelayer_norm_epsposition_embedding_type quant_mode force_dequant)selfrrrrr rr!r"r#r$r%r&rrrr'r(r)kwargs __class__Q/tmp/pip-unpacked-wheel-zw5xktn0/transformers/models/ibert/configuration_ibert.pyrVs zIBertConfig.__init__)r r r r r r rrrrrrrrrrFr)__name__ __module__ __qualname____doc__Z model_typer __classcell__r.r.r,r/rs*5rc@s.eZdZeeeeeeffdddZdS)IBertOnnxConfig)returncCs6|jdkrdddd}n ddd}td|fd|fgS) Nzmultiple-choicebatchchoicesequence)rrr)rrZ input_idsZattention_mask)Ztaskr)r*Z dynamic_axisr.r.r/inputss  zIBertOnnxConfig.inputsN)r0r1r2propertyrstrintr:r.r.r.r/r5sr5N)r3 collectionsrtypingrZconfiguration_utilsrZonnxrutilsrZ get_loggerr0loggerrr5r.r.r.r/s      a