xmodaler.modeling.meta_arch

xmodaler.modeling.meta_arch.build_model(cfg)

Build the whole model architecture, defined by cfg.MODEL.META_ARCHITECTURE. Note that it does not load any weights from cfg.

xmodaler.modeling.meta_arch.add_config(cfg, tmp_cfg)

class xmodaler.modeling.meta_arch.base_enc_dec.BaseEncoderDecoder(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher)

Bases: Module

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

classmethod add_config(cfg, tmp_cfg)

bind_or_init_weights()

decode_beam_search(batched_inputs, output_sents=False)

forward(batched_inputs, use_beam_search=None, output_sents=False)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

classmethod from_config(cfg)

abstract get_extended_attention_mask(batched_inputs)

greedy_decode(batched_inputs, output_sents=False)

preprocess_batch(batched_inputs)

training: bool

class xmodaler.modeling.meta_arch.RnnAttEncoderDecoder(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher)

Bases: BaseEncoderDecoder

get_extended_attention_mask(batched_inputs)

training: bool

class xmodaler.modeling.meta_arch.TransformerEncoderDecoder(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor)

Bases: BaseEncoderDecoder

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

classmethod from_config(cfg)

get_extended_attention_mask(batched_inputs)

training: bool

class xmodaler.modeling.meta_arch.TDENBiTransformer(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor)

Bases: TransformerEncoderDecoder

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

classmethod from_config(cfg)

get_extended_attention_mask(batched_inputs)

training: bool

class xmodaler.modeling.meta_arch.TDENPretrain(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor, similarity_predictor)

Bases: TransformerEncoderDecoder

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor, similarity_predictor)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

classmethod add_config(cfg, tmp_cfg)

classmethod from_config(cfg)

training: bool

class xmodaler.modeling.meta_arch.TDENCaptioner(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor)

Bases: TransformerEncoderDecoder

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

classmethod from_config(cfg)

training: bool

class xmodaler.modeling.meta_arch.UniterPretrain(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor, v_regressor, itm_predictor, tasks, mix_ratio)

Bases: TransformerEncoderDecoder

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor, v_regressor, itm_predictor, tasks, mix_ratio)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

classmethod add_config(cfg, tmp_cfg)

classmethod from_config(cfg)

preprocess_inputs(inputs, task_name)

training: bool

class xmodaler.modeling.meta_arch.UniterForMMUnderstanding(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor, itm_predictor)

Bases: TransformerEncoderDecoder

__init__(*, vocab_size, max_seq_len, token_embed, visual_embed, encoder, decoder, predictor, greedy_decoder, beam_searcher, v_predictor, itm_predictor)

Initializes internal Module state, shared by both nn.Module and ScriptModule.

bind_or_init_weights()

classmethod from_config(cfg)

get_extended_attention_mask(batched_inputs)

training: bool