hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config module

class hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config.DecoderConfig(add_cross_attention: bool = True, vocab_size: int = 42001, attn_pdrop: float = 0.1, bos_token_id: int = 5, embd_pdrop: float = 0.1, eos_token_id: int = 5, gradient_checkpointing: bool = False, initializer_range: float = 0.02, layer_norm_epsilon: float = 1e-05, model_type: str = 'gpt2', n_ctx: int = 1024, n_embd: int = 768, n_head: int = 12, n_inner: int = None, n_layer: int = 12, n_positions: int = 1024, resid_pdrop: float = 0.1, summary_activation: bool = False, summary_first_dropout: float = 0.1, use_cache: bool = True)

Bases: ModelConfig

add_cross_attention: bool = True

attn_pdrop: float = 0.1

bos_token_id: int = 5

embd_pdrop: float = 0.1

eos_token_id: int = 5

gradient_checkpointing: bool = False

initializer_range: float = 0.02

layer_norm_epsilon: float = 1e-05

model_type: str = 'gpt2'

n_ctx: int = 1024

n_embd: int = 768

n_head: int = 12

n_inner: int = None

n_layer: int = 12

n_positions: int = 1024

name: str = 'vit_gpt2_decoder'

resid_pdrop: float = 0.1

summary_activation: bool = False

summary_first_dropout: float = 0.1

use_cache: bool = True

vocab_size: int = 42001

class hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config.EncoderConfig(hidden_size: int = 768, num_hidden_layers: int = 12, num_attention_heads: int = 12, intermediate_size: int = 3072, hidden_act: str = 'gelu', hidden_dropout_prob: float = 0.0, attention_probs_dropout_prob: float = 0.0, initializer_range: float = 0.02, layer_norm_eps: float = 1e-12, image_size: int = 224, patch_size: int = 16, num_channels: int = 3, qkv_bias: bool = True, encoder_stride: int = 16)

Bases: ModelConfig

attention_probs_dropout_prob: float = 0.0

encoder_stride: int = 16

hidden_act: str = 'gelu'

hidden_dropout_prob: float = 0.0

hidden_size: int = 768

image_size: int = 224

initializer_range: float = 0.02

intermediate_size: int = 3072

layer_norm_eps: float = 1e-12

name: str = 'vit_gpt2_encoder'

num_attention_heads: int = 12

num_channels: int = 3

num_hidden_layers: int = 12

patch_size: int = 16

qkv_bias: bool = True

class hezar.models.image2text.vit_gpt2.vit_gpt2_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_new_tokens: int = 24, no_repeat_ngram_size: int = 3, num_beams: int = 4, pad_token_id: int = 1)

Bases: ModelConfig

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_new_tokens: int = 24

no_repeat_ngram_size: int = 3

num_beams: int = 4

pad_token_id: int = 1

class hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config.ViTGPT2Image2TextConfig(encoder: hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config.EncoderConfig = <factory>, decoder: hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config.DecoderConfig = <factory>, generation: hezar.models.image2text.vit_gpt2.vit_gpt2_image2text_config.GenerationConfig = <factory>)

Bases: ModelConfig

decoder: DecoderConfig

encoder: EncoderConfig

generation: GenerationConfig

name: str = 'vit_gpt2_image2text'