Source code for transformers.configuration_bart

# coding=utf-8
# Copyright 2020 The Fairseq Authors and The HuggingFace Inc. team.
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# See the License for the specific language governing permissions and
# limitations under the License.
""" BART configuration """

import logging

from .configuration_utils import PretrainedConfig

logger = logging.getLogger(__name__)

    "facebook/bart-base": "",
    "facebook/bart-large": "",
    "facebook/bart-large-mnli": "",
    "facebook/bart-large-cnn": "",
    "facebook/bart-large-xsum": "",
    "facebook/mbart-large-en-ro": "",
    "yjernite/bart_eli5": "",

[docs]class BartConfig(PretrainedConfig): r""" Configuration class for Bart. Parameters are renamed from the fairseq implementation """ model_type = "bart" def __init__( self, activation_dropout=0.0, extra_pos_embeddings=2, activation_function="gelu", vocab_size=50265, d_model=1024, encoder_ffn_dim=4096, encoder_layers=12, encoder_attention_heads=16, decoder_ffn_dim=4096, decoder_layers=12, decoder_attention_heads=16, encoder_layerdrop=0.0, decoder_layerdrop=0.0, attention_dropout=0.0, dropout=0.1, max_position_embeddings=1024, init_std=0.02, classifier_dropout=0.0, num_labels=3, is_encoder_decoder=True, pad_token_id=1, bos_token_id=0, eos_token_id=2, normalize_before=False, add_final_layer_norm=False, scale_embedding=False, normalize_embedding=True, static_position_embeddings=False, add_bias_logits=False, **common_kwargs ): r""" :class:`~transformers.BartConfig` is the configuration class for `BartModel`. Examples:: >>> from transformers import BartConfig, BartModel >>> config = BartConfig.from_pretrained('facebook/bart-large') >>> model = BartModel(config) """ if "hidden_size" in common_kwargs: raise ValueError("hidden size is called d_model") super().__init__( num_labels=num_labels, pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, is_encoder_decoder=is_encoder_decoder, **common_kwargs, ) self.vocab_size = vocab_size self.d_model = d_model # encoder_embed_dim and decoder_embed_dim self.encoder_ffn_dim = encoder_ffn_dim self.encoder_layers = self.num_hidden_layers = encoder_layers self.encoder_attention_heads = encoder_attention_heads self.encoder_layerdrop = encoder_layerdrop self.decoder_layerdrop = decoder_layerdrop self.decoder_ffn_dim = decoder_ffn_dim self.decoder_layers = decoder_layers self.decoder_attention_heads = decoder_attention_heads self.max_position_embeddings = max_position_embeddings self.init_std = init_std # Normal(0, this parameter) self.activation_function = activation_function # Params introduced for Mbart self.scale_embedding = scale_embedding # scale factor will be sqrt(d_model) if True self.normalize_embedding = normalize_embedding # True for mbart, False otherwise self.normalize_before = normalize_before # combo of fairseq's encoder_ and decoder_normalize_before self.add_final_layer_norm = add_final_layer_norm # Params introduced for Marian self.add_bias_logits = add_bias_logits self.static_position_embeddings = static_position_embeddings # 3 Types of Dropout self.attention_dropout = attention_dropout self.activation_dropout = activation_dropout self.dropout = dropout # Classifier stuff self.classif_dropout = classifier_dropout # pos embedding offset self.extra_pos_embeddings = self.pad_token_id + 1 @property def num_attention_heads(self) -> int: return self.encoder_attention_heads @property def hidden_size(self) -> int: return self.d_model
[docs] def is_valid_mbart(self) -> bool: """Is the configuration aligned with the MBART paper.""" if self.normalize_before and self.add_final_layer_norm and self.scale_embedding: return True if self.normalize_before or self.add_final_layer_norm or self.scale_embedding:"This configuration is a mixture of MBART and BART settings") return False
class MBartConfig(BartConfig): model_type = "mbart"