Functions — github.com/airsplay/lxmert

hub / github.com/airsplay/lxmert / functions

Functions214 in github.com/airsplay/lxmert

⨍ Functions 214◇ Types & classes 53

↓ 21 callersMethod__init__

(self)

src/lxrt/modeling.py:127

↓ 12 callersMethodload

(self, path)

src/tasks/gqa.py:167

↓ 6 callersFunctionload_obj_tsv

Load object features from tsv file. :param fname: The path to the tsv file. :param topk: Only load features for top K images (lines) in the t

src/utils.py:16

↓ 5 callersMethoddim

(self)

src/lxrt/entry.py:106

↓ 5 callersMethodload

(self, path)

src/pretrain/lxmert_pretrain.py:394

↓ 4 callersMethodevaluate

(self, eval_tuple: DataTuple, dump=None)

src/tasks/gqa.py:147

↓ 4 callersMethodevaluate

Evaluate all data in data_tuple.

src/tasks/vqa.py:150

↓ 4 callersMethodfrom_pretrained

Instantiate a PreTrainedBertModel from a pre-trained model file. Download and cache the pre-trained model file if needed.

src/lxrt/tokenization.py:136

↓ 4 callersFunctionget_data_tuple

(splits: str, bs:int, shuffle=False, drop_last=False)

src/tasks/vqa.py:20

↓ 4 callersFunctionget_tuple

(splits: str, bs:int, shuffle=False, drop_last=False)

src/tasks/gqa.py:21

↓ 4 callersFunctionget_tuple

(splits: str, bs:int, shuffle=False, drop_last=False)

src/tasks/nlvr2.py:19

↓ 4 callersMethodstep

Performs a single optimization step. Arguments: closure (callable, optional): A closure that reevaluates the model

src/lxrt/optimization.py:101

↓ 3 callersFunctioncached_path

Given something that might be a URL (or might be a local path), determine which. If it's a URL, download the file and cache it, and retur

src/lxrt/file_utils.py:81

↓ 3 callersMethoddump_result

Dump the result to a GQA-challenge submittable json file. GQA json file submission requirement: results = [result]

src/tasks/gqa_data.py:171

↓ 3 callersMethodevaluate

(self, eval_tuple: DataTuple, dump=None)

src/tasks/nlvr2.py:133

↓ 3 callersMethodmulti_gpu

(self)

src/lxrt/entry.py:102

↓ 3 callersMethodtranspose_for_scores

(self, x)

src/lxrt/modeling.py:315

↓ 3 callersFunctionwhitespace_tokenize

Runs basic whitespace cleaning and splitting on a piece of text.

src/lxrt/tokenization.py:63

↓ 2 callersMethodans2id

(self, ans)

src/pretrain/qa_answer_table.py:61

↓ 2 callersMethodconvert_ans

(self, ans)

src/pretrain/qa_answer_table.py:45

↓ 2 callersMethodconvert_tokens_to_ids

Converts a sequence of tokens into ids using the vocab.

src/lxrt/tokenization.py:115

↓ 2 callersMethodevaluate

(self, uid2ans: dict, pprint=False)

src/pretrain/lxmert_data.py:224

↓ 2 callersMethodforward

(self, examples)

src/pretrain/lxmert_pretrain.py:251

↓ 2 callersFunctionget_tuple

(splits: str, bs: int, shuffle=False, drop_last=False, topk=-1)

src/pretrain/lxmert_pretrain.py:23

↓ 2 callersMethodid2ans

(self, ans_id)

src/pretrain/qa_answer_table.py:64

↓ 2 callersMethodload

(self, path)

src/tasks/nlvr2.py:142

↓ 2 callersMethodload

(self, path)

src/tasks/vqa.py:170

↓ 2 callersMethodload_data

(self, name, number)

src/tasks/gqa_data.py:67

↓ 2 callersFunctionload_lxmert_qa

Load model weights from LXMERT pre-training. The answers in the fine-tuned QA task (indicated by label2ans) would also be properly in

src/pretrain/qa_answer_table.py:84

↓ 2 callersFunctionmake_uid

(img_id, dset, sent_idx)

src/pretrain/lxmert_data.py:78

↓ 2 callersMethodpredict

(self, eval_tuple: DataTuple, dump=None)

src/tasks/gqa.py:130

↓ 2 callersMethodpredict

(self, eval_tuple: DataTuple, dump=None)

src/tasks/nlvr2.py:117

↓ 2 callersMethodpredict

Predict the answers to questions in a data split. :param eval_tuple: The data tuple to be evaluated. :param dump: The path o

src/tasks/vqa.py:126

↓ 2 callersMethodsave

(self, name)

src/tasks/gqa.py:163

↓ 2 callersMethodsave

(self, name)

src/tasks/nlvr2.py:138

↓ 2 callersMethodsave

(self, name)

src/tasks/vqa.py:166

↓ 2 callersMethodsave

(self, name)

src/pretrain/lxmert_pretrain.py:390

↓ 2 callersFunctionset_visual_config

(args)

src/lxrt/entry.py:74

↓ 2 callersFunctionsplit_s3_path

Split a full s3 path into the bucket name and path.

src/lxrt/file_utils.py:111

↓ 2 callersMethodtokenize

(self, text)

src/lxrt/tokenization.py:105

↓ 2 callersMethodused

(self, ans)

src/pretrain/qa_answer_table.py:73

↓ 1 callersMethod__init__

(self, splits: str)

src/tasks/nlvr2_data.py:32

↓ 1 callersMethod__init__

(self, splits: str)

src/tasks/gqa_data.py:32

↓ 1 callersMethod__init__

(self, splits: str)

src/tasks/vqa_data.py:47

↓ 1 callersMethod__init__

(self, uid, sent, visual_feats=None, obj_labels=None, attr_labels=None, is_m

src/pretrain/lxmert_data.py:28

↓ 1 callersMethod_clean_text

Performs invalid character removal and whitespace cleanup on text.

src/lxrt/tokenization.py:277

↓ 1 callersMethod_is_chinese_char

Checks whether CP is the codepoint of a CJK character.

src/lxrt/tokenization.py:255

↓ 1 callersFunction_is_control

Checks whether `chars` is a control character.

src/lxrt/tokenization.py:363

↓ 1 callersFunction_is_punctuation

Checks whether `chars` is a punctuation character.

src/lxrt/tokenization.py:375

↓ 1 callersFunction_is_whitespace

Checks whether `chars` is a whitespace character.

src/lxrt/tokenization.py:351

↓ 1 callersMethod_run_split_on_punc

Splits punctuation on a piece of text.

src/lxrt/tokenization.py:220

↓ 1 callersMethod_run_strip_accents

Strips accents from a piece of text.

src/lxrt/tokenization.py:209

↓ 1 callersMethod_tokenize_chinese_chars

Adds whitespace around any CJK character.

src/lxrt/tokenization.py:242

↓ 1 callersMethodans2id_map

(self)

src/pretrain/qa_answer_table.py:67

↓ 1 callersFunctionconvert_example_to_features

Convert a raw sample (pair of sentences as tokenized strings) into a proper training sample with IDs, LM labels, input_mask, CLS and SEP toke

src/pretrain/lxmert_pretrain.py:134

↓ 1 callersFunctionconvert_sents_to_features

Loads a data file into a list of `InputBatch`s.

src/lxrt/entry.py:36

↓ 1 callersMethodcross_att

(self, lang_input, lang_attention_mask, visn_input, visn_attention_mask)

src/lxrt/modeling.py:455

↓ 1 callersMethodevaluate_epoch

(self, eval_tuple: DataTuple, iters: int=-1)

src/pretrain/lxmert_pretrain.py:360

↓ 1 callersMethodfrom_dict

Constructs a `BertConfig` from a Python dictionary of parameters.

src/lxrt/modeling.py:234

↓ 1 callersMethodfrom_json_file

Constructs a `BertConfig` from a json file of parameters.

src/lxrt/modeling.py:242

↓ 1 callersFunctiongelu

Implementation of the gelu activation function. For information: OpenAI GPT's gelu is slightly different (and gives slightly different results

src/lxrt/modeling.py:112

↓ 1 callersFunctiongenerate_tsv

(prototxt, weights, image_ids, outfile)

data/mscoco_imgfeat/extract_coco_image.py:54

↓ 1 callersFunctiongenerate_tsv

(prototxt, weights, image_ids, outfile)

data/nlvr2_imgfeat/extract_nlvr2_image.py:69

↓ 1 callersFunctiongenerate_tsv

(prototxt, weights, image_ids, outfile)

data/vg_gqa_imgfeat/extract_gqa_image.py:45

↓ 1 callersFunctionget_detections_from_im

:param net: :param im_file: full path to an image :param image_id: :param conf_thresh: :return: all information from detection an

data/mscoco_imgfeat/extract_coco_image.py:83

↓ 1 callersFunctionget_detections_from_im

:param net: :param im_file: full path to an image :param image_id: :param conf_thresh: :return: all information from detection an

data/nlvr2_imgfeat/extract_nlvr2_image.py:98

↓ 1 callersFunctionget_detections_from_im

:param net: :param im_file: full path to an image :param image_id: :param conf_thresh: :return: all information from detection an

data/vg_gqa_imgfeat/extract_gqa_image.py:74

↓ 1 callersFunctionget_from_cache

Given a URL, look for the corresponding dataset in the local cache. If it's not there, download it. Then return the path to the cached file.

src/lxrt/file_utils.py:172

↓ 1 callersFunctionget_optimizer

(optim)

src/param.py:11

↓ 1 callersFunctionhttp_get

(url, temp_file)

src/lxrt/file_utils.py:160

↓ 1 callersMethodload

(module, prefix='')

src/lxrt/modeling.py:812

↓ 1 callersMethodload

(self, path)

src/lxrt/entry.py:126

↓ 1 callersFunctionload_image_ids

images in the same directory are in the same split

data/mscoco_imgfeat/extract_coco_image.py:42

↓ 1 callersFunctionload_image_ids

images in the same directory are in the same sequential region, but with no internal ordering

data/nlvr2_imgfeat/extract_nlvr2_image.py:45

↓ 1 callersFunctionload_image_ids

(img_root)

data/vg_gqa_imgfeat/extract_gqa_image.py:35

↓ 1 callersMethodload_lxmert

(self, path)

src/pretrain/lxmert_pretrain.py:399

↓ 1 callersFunctionload_tf_weights_in_bert

Load tf checkpoints in a pytorch model

src/lxrt/modeling.py:51

↓ 1 callersFunctionload_vocab

Loads a vocabulary file into a dictionary.

src/lxrt/tokenization.py:48

↓ 1 callersMethodoracle_score

(data_tuple)

src/tasks/gqa.py:153

↓ 1 callersMethodoracle_score

(data_tuple)

src/tasks/vqa.py:156

↓ 1 callersMethodoutput_fc

(self, lang_input, visn_input)

src/lxrt/modeling.py:467

↓ 1 callersFunctionparse_args

Parse input arguments

data/mscoco_imgfeat/extract_coco_image.py:137

↓ 1 callersFunctionparse_args

Parse input arguments

data/nlvr2_imgfeat/extract_nlvr2_image.py:152

↓ 1 callersFunctionparse_args

Parse input arguments

data/vg_gqa_imgfeat/extract_gqa_image.py:128

↓ 1 callersFunctionparse_args

()

src/param.py:33

↓ 1 callersFunctionrandom_feat

(feats)

src/pretrain/lxmert_pretrain.py:110

↓ 1 callersMethodrandom_feat

Get a random obj feat from the dataset.

src/pretrain/lxmert_data.py:136

↓ 1 callersFunctionrandom_word

Masking some random tokens for Language Model task with probabilities as in the original BERT paper. :param tokens: list of str, tokenized se

src/pretrain/lxmert_pretrain.py:71

↓ 1 callersFunctions3_etag

Check ETag on S3 object.

src/lxrt/file_utils.py:144

↓ 1 callersFunctions3_get

Pull a file directly from S3.

src/lxrt/file_utils.py:153

↓ 1 callersMethodself_att

(self, lang_input, lang_attention_mask, visn_input, visn_attention_mask)

src/lxrt/modeling.py:461

↓ 1 callersMethodto_dict

Serializes this instance to a Python dictionary.

src/lxrt/modeling.py:251

↓ 1 callersMethodto_json_string

Serializes this instance to a JSON string.

src/lxrt/modeling.py:256

↓ 1 callersMethodtrain

(self, train_tuple, eval_tuple)

src/tasks/gqa.py:80

↓ 1 callersMethodtrain

(self, train_tuple, eval_tuple)

src/tasks/nlvr2.py:74

↓ 1 callersMethodtrain

(self, train_tuple, eval_tuple)

src/tasks/vqa.py:80

↓ 1 callersMethodtrain

(self, train_tuple: DataTuple, eval_tuple: DataTuple)

src/pretrain/lxmert_pretrain.py:309

↓ 1 callersMethodtrain_batch

(self, optim, batch)

src/pretrain/lxmert_pretrain.py:289

↓ 1 callersFunctionurl_to_filename

Convert `url` into a hashed filename in a repeatable way. If `etag` is specified, append its hash to the url's, delimited by a period.

src/lxrt/file_utils.py:37

↓ 1 callersMethodvalid_batch

(self, batch)

src/pretrain/lxmert_pretrain.py:301

next →1–100 of 214, ranked by callers