hezar.models.image2text.beit_roberta.beit_roberta_image2text_config module

class hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.BeitRobertaImage2TextConfig(encoder: hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.EncoderConfig = <factory>, decoder: hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.DecoderConfig = <factory>, generation: hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.GenerationConfig = <factory>)

Bases: ModelConfig

decoder: DecoderConfig

encoder: EncoderConfig

generation: GenerationConfig

name: str = 'beit_roberta_image2text'

class hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.DecoderConfig(is_decoder: bool = True, add_cross_attention: bool = True, attention_probs_dropout_prob: float = 0.1, bos_token_id: int = 0, eos_token_id: int = 2, classifier_dropout: float = None, gradient_checkpointing: bool = False, hidden_act: str = 'gelu', hidden_dropout_prob: float = 0.1, hidden_size: int = 768, initializer_range: int = 0.02, intermediate_size: int = 3072, layer_norm_eps: float = 1e-12, max_position_embeddings: int = 514, num_attention_heads: int = 12, num_hidden_layers: int = 12, pad_token_id: int = 1, position_embedding_type: str = 'absolute', type_vocab_size: int = 1, use_cache: bool = True, vocab_size: int = 42000)

Bases: ModelConfig

add_cross_attention: bool = True

attention_probs_dropout_prob: float = 0.1

bos_token_id: int = 0

classifier_dropout: float = None

eos_token_id: int = 2

gradient_checkpointing: bool = False

hidden_act: str = 'gelu'

hidden_dropout_prob: float = 0.1

hidden_size: int = 768

initializer_range: int = 0.02

intermediate_size: int = 3072

is_decoder: bool = True

layer_norm_eps: float = 1e-12

max_position_embeddings: int = 514

name: str = 'beit_roberta_decoder'

num_attention_heads: int = 12

num_hidden_layers: int = 12

pad_token_id: int = 1

position_embedding_type: str = 'absolute'

type_vocab_size: int = 1

use_cache: bool = True

vocab_size: int = 42000

class hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.EncoderConfig

Bases: ModelConfig

attention_probs_dropout_prob = 0.0

auxiliary_channels = 256

auxiliary_concat_input = False

auxiliary_loss_weight = 0.4

auxiliary_num_convs = 1

drop_path_rate = 0.1

hidden_act = 'gelu'

hidden_dropout_prob = 0.0

hidden_size = 768

image_size = 224

initializer_range = 0.02

intermediate_size = 3072

layer_norm_eps = 1e-12

layer_scale_init_value = 0.1

name: str = 'beit_roberta_encoder'

num_attention_heads = 12

num_channels = 3

num_hidden_layers = 12

out_indices = [3, 5, 7, 11]

patch_size = 16

pool_scales = [1, 2, 3, 6]

semantic_loss_ignore_index = 255

use_absolute_position_embeddings = False

use_auxiliary_head = True

use_mask_token = False

use_mean_pooling = True

use_relative_position_bias = False

use_shared_relative_position_bias = False

vocab_size = 8192

class hezar.models.image2text.beit_roberta.beit_roberta_image2text_config.GenerationConfig(bos_token_id: int = 0, decoder_start_token_id: int = 0, early_stopping: bool = True, eos_token_id: int = 2, length_penalty: float = 2.0, max_length: int = 64, no_repeat_ngram_size: int = 3, num_beams: int = 4, pad_token_id: int = 1)

Bases: object

bos_token_id: int = 0

decoder_start_token_id: int = 0

early_stopping: bool = True

eos_token_id: int = 2

length_penalty: float = 2.0

max_length: int = 64

no_repeat_ngram_size: int = 3

num_beams: int = 4

pad_token_id: int = 1