Functions — github.com/NUSTM/VLP-MABSA

hub / github.com/NUSTM/VLP-MABSA / functions

Functions191 in github.com/NUSTM/VLP-MABSA

⨍ Functions 191◇ Types & classes 41 ↳ Endpoints 1

↓ 75 callersMethodinfo

(self, message, pad=False)

src/utils.py:50

↓ 33 callersMethodconvert_tokens_to_ids

(self, tokens)

src/data/tokenization_new.py:667

↓ 21 callersMethodload

(module: nn.Module, prefix="")

src/model/mixins.py:789

↓ 13 callersMethod__init__

( self, embed_dim, num_heads, dropout=0.0, bias=Tr

src/model/modeling_bart.py:676

↓ 13 callersMethodfrom_pretrained

r"""Instantiate a pretrained pytorch model from a pre-trained model configuration. The model is set in evaluation mode by default using model

src/model/mixins.py:552

↓ 9 callersFunctionLayerNorm

(normalized_shape, eps=1e-5, elementwise_affine=True)

src/model/modeling_bart.py:1069

↓ 7 callersMethod__init__

(self)

src/model/modules.py:326

↓ 5 callersMethoddecode

(self, tokens, state, only_sc=False)

src/model/modules.py:321

↓ 4 callersFunction_compute_f_pre_rec

r""" :param tp: int, true positive :param fn: int, false negative :param fp: int, false positive :return: (f, pre, rec)

src/model/metrics.py:196

↓ 4 callersMethodadd

Add a new hypothesis to the list.

src/model/generater.py:765

↓ 4 callersMethodencode

(self, *args, **kwargs)

src/data/tokenization_new.py:159

↓ 4 callersMethodforward

Input shape: Time(SeqLen) x Batch x Channel

src/model/modeling_bart.py:703

↓ 4 callersFunctioninvert_mask

Turns 1->0, 0->1, False->True, True-> False

src/model/modeling_bart.py:134

↓ 4 callersMethodpad_tokens

(self, tokens)

src/data/tokenization_new.py:162

↓ 3 callersFunction_compute_tp_fn_fp

(ps, ts)

src/model/metrics.py:211

↓ 3 callersMethod_init_weights

(self, module)

src/model/modeling_bart.py:173

↓ 3 callersFunction_make_linear_from_emb

(emb)

src/model/modeling_bart.py:198

↓ 3 callersFunction_prepare_bart_decoder_inputs

Prepare masks that ignore padding tokens in the decoder and a causal mask for the decoder if none are provided. This mimics the default behavior i

src/model/modeling_bart.py:140

↓ 3 callersMethod_shape

(self, tensor, seq_len, bsz)

src/model/modeling_bart.py:699

↓ 3 callersFunctionfine_tune

(epoch, model, train_loader, test_loader, metric,

src/training.py:105

↓ 3 callersMethodforward

(self, tgt_tokens, pred, mask)

src/model/modules.py:331

↓ 2 callersMethod_force_token_ids_generation

force one of token_ids to be generated by setting prob of all other tokens to 0 (logprob=-float("inf"))

src/model/modeling_bart.py:1381

↓ 2 callersMethod_force_token_ids_generation

force one of token_ids to be generated by setting prob of all other tokens to 0

src/model/mixins.py:407

↓ 2 callersMethod_init_weights

(self, module)

src/model/modules.py:484

↓ 2 callersFunction_reorder_buffer

(attn_cache, new_order)

src/model/modeling_bart.py:667

↓ 2 callersMethodgenerate

:param State state: encoder结果的State, 是与Decoder配套是用的 :param torch.LongTensor,None tokens: batch_size x length, 开始的token :retu

src/model/generater.py:209

↓ 2 callersMethodget_base_tokenizer

(self)

src/data/tokenization_new.py:673

↓ 2 callersMethodget_encoder

(self)

src/model/mixins.py:436

↓ 2 callersMethodprepare_state

(self, input_ids, image_features, attention_

src/model/MAESC_model.py:100

↓ 2 callersMethodreorder_state

(self, indices: torch.LongTensor)

src/model/model.py:220

↓ 2 callersMethodresize_token_embeddings

(self, new_num_tokens: int)

src/model/mixins.py:442

↓ 2 callersFunctionshift_tokens_right

Shift input ids one token to the right, and wrap the last non pad token (usually <eos>).

src/model/modeling_bart.py:212

↓ 1 callersMethod__init__

(self, eos_token_id, num_labels, opinion_first=True)

src/model/metrics.py:242

↓ 1 callersMethod__init__

Initialize n-best list of hypotheses.

src/model/generater.py:748

↓ 1 callersMethod__init__

(self, encoder_output, encoder_mask, src_tokens, first, src_embed_outputs)

src/model/MAESC_model.py:150

↓ 1 callersMethod__init__

(self, encoder_output, encoder_mask, src_tokens, first, src_embed_outputs)

src/model/model.py:212

↓ 1 callersFunction_beam_search_generate

(decoder: Seq2SeqDecoder, tokens=None, state=None,

src/model/generater.py:499

↓ 1 callersMethod_embed_multi_modal

embed textual and visual inputs and combine them into one embedding

src/model/modules.py:81

↓ 1 callersMethod_get_logger

get a logger for displaying information to console or log to file (optional) :param log_dir: str, logging path. None for not log to f

src/utils.py:60

↓ 1 callersMethod_init_weight

Identical to the XLM create_sinusoidal_embeddings except features are not interleaved. The cos features are in the 2nd half of the vector. [di

src/model/modeling_bart.py:1613

↓ 1 callersMethod_load_from_state_dict

r"""Copies parameters and buffers from :attr:`state_dict` into only this module, but not its descendants. This is called on every submodule

src/model/mixins.py:463

↓ 1 callersMethod_mask_tokens

Prepare masked tokens inputs/labels for masked language modeling: 80% MASK, 10% random, 10% original. :param inputs: torch.LongTenso

src/data/collation.py:185

↓ 1 callersFunction_no_beam_search_generate

(decoder: Seq2SeqDecoder, state, tokens=None,

src/model/generater.py:300

↓ 1 callersMethod_pad_message

(self, message)

src/utils.py:47

↓ 1 callersMethod_resize_final_logits_bias

(self, new_num_tokens: int, old_num_tokens: int)

src/model/modeling_bart.py:1239

↓ 1 callersMethod_resize_final_logits_bias

(self, new_num_tokens: int, old_num_tokens: int)

src/model/mixins.py:449

↓ 1 callersMethod_use_saved_state

(self, k, v, saved_state, key_padding_mask, static_kv, bsz)

src/model/modeling_bart.py:801

↓ 1 callersMethodbuild_model

(self, args, bart_model, tokenizer,

src/model/MAESC_model.py:19

↓ 1 callersMethodbuild_model

(self, args, bart_model, tokenizer,

src/model/model.py:31

↓ 1 callersFunctioncleanup_process

()

src/utils.py:16

↓ 1 callersMethodcreate_id2idx

(self)

src/data/dataset.py:37

↓ 1 callersMethoddecode

(self, token_ids, skip_special_tokens=False)

src/data/tokenization_new.py:663

↓ 1 callersMethodencode_ae_oe

(self, label, aspect_spans, opinion_spans)

src/data/tokenization_new.py:448

↓ 1 callersMethodencode_aesc

(self, label, aesc_spans, aesc_max_len)

src/data/tokenization_new.py:393

↓ 1 callersMethodencode_anp_generate

(self, ANP_words)

src/data/tokenization_new.py:350

↓ 1 callersMethodencode_condition

tokenize text, image features and event the output format (after decoded back): task_type [<img> <img_feat> ... <img_feat> </

src/data/tokenization_new.py:186

↓ 1 callersMethodencode_label

(self, label, img_num=None)

src/data/tokenization_new.py:268

↓ 1 callersMethodencode_mrm

(self, box_cls)

src/data/tokenization_new.py:510

↓ 1 callersMethodencode_senti

(self, sentis)

src/data/tokenization_new.py:313

↓ 1 callersMethodencode_twitter_ae

(self, label, aspect_spans, ae_max_len)

src/data/tokenization_new.py:529

↓ 1 callersMethodencode_twitter_sc

(self, label, aesc_spans, aesc_max_len)

src/data/tokenization_new.py:586

↓ 1 callersMethodevaluate

(self, oe_target_span, pred, tgt_tokens)

src/model/metrics.py:257

↓ 1 callersFunctionfill_with_neg_inf

FP16-compatible function that fills a input_ids with -inf.

src/model/modeling_bart.py:1080

↓ 1 callersMethodget_ANP_word

(self, distribution)

src/data/dataset.py:60

↓ 1 callersMethodget_aesc_spans

(self, dic)

src/data/dataset.py:155

↓ 1 callersMethodget_aspect_spans

(self, id)

src/data/dataset.py:81

↓ 1 callersMethodget_cls

(self, id)

src/data/dataset.py:89

↓ 1 callersMethodget_gt_aspect_senti

(self, dic)

src/data/dataset.py:161

↓ 1 callersMethodget_img_ANP

(self, idx)

src/data/dataset.py:65

↓ 1 callersMethodget_img_region_box

(self, id)

src/data/dataset.py:46

↓ 1 callersMethodget_img_region_box

(self, id)

src/data/dataset.py:146

↓ 1 callersMethodget_metric

(self, reset=True)

src/model/metrics.py:315

↓ 1 callersMethodget_opinion_spans

(self, id)

src/data/dataset.py:85

↓ 1 callersMethodget_sentence

(self, id)

src/data/dataset.py:76

↓ 1 callersMethodget_sentiment

(self, id)

src/data/dataset.py:72

↓ 1 callersMethodis_done

If there are enough hypotheses and that none of the hypotheses being generated can become better than the worst one in the heap, then

src/model/generater.py:781

↓ 1 callersFunctionmain

(rank, args)

pretrain.py:26

↓ 1 callersFunctionmain

(rank, args)

twitter_sc_training.py:30

↓ 1 callersFunctionmain

(rank, args)

MAESC_training.py:30

↓ 1 callersFunctionmain

(rank, args)

twitter_ae_training.py:30

↓ 1 callersFunctionmake_padding_mask

True for pad tokens

src/model/modeling_bart.py:221

↓ 1 callersFunctionparse_args

()

pretrain.py:245

↓ 1 callersFunctionparse_args

()

twitter_sc_training.py:298

↓ 1 callersFunctionparse_args

()

MAESC_training.py:260

↓ 1 callersFunctionparse_args

()

twitter_ae_training.py:274

↓ 1 callersMethodpredict

给定source的内容，输出generate的内容 :param torch.LongTensor src_tokens: bsz x max_len :param torch.LongTensor src_seq_len: bsz

src/model/generater.py:84

↓ 1 callersMethodprepare_state

(self, input_ids, image_features, attention_

src/model/model.py:106

↓ 1 callersFunctionpretrain

(task_list, epoch, model, train_loaders, optimizer_dict,

src/training.py:10

↓ 1 callersMethodprocess_ANP_distribution

(self, distribution)

src/data/dataset.py:53

↓ 1 callersFunctionsave_training_data

(path, optimizer=None, scaler=None, epoch=None)

src/utils.py:20

↓ 1 callersFunctionsc_generate

(decoder: Seq2SeqDecoder, state, tokens=None, gt_tokens=None,

src/model/generater.py:413

Method__call__

(self, batch)

src/data/collation.py:82

Method__getitem__

(self, index)

src/data/dataset.py:96

Method__getitem__

(self, index)

src/data/dataset.py:167

Method__init__

(self, log_dir=None, enabled=True, pad_length=50)

src/utils.py:43

Method__init__

( self, activation_dropout=0.0, extra_pos_embeddings=2, activa

src/model/config.py:5

Method__init__

(self, config: BartConfig)

src/model/modeling_bart.py:233

Method__init__

(self, config: BartConfig, embed_tokens)

src/model/modeling_bart.py:296

Method__init__

(self, config: BartConfig)

src/model/modeling_bart.py:400

Method__init__

(self, config: BartConfig, embed_tokens: nn.Embedding)

src/model/modeling_bart.py:499

next →1–100 of 191, ranked by callers