(data, pad_size=200, )
| 63 | |
| 64 | |
| 65 | def gen_bert_vector(data, pad_size=200, ): |
| 66 | model = Bert('./models/pytorch_pretrained_bert/bert_pretrain/', './temp/', load_pretrained_bert=True, |
| 67 | bert_config=None) |
| 68 | b_data = bert.pre_process(data, tgt=[list('NONE')], oracle_ids=[0], flag_i=0) |
| 69 | indexed_tokens, labels, segments_ids, cls_ids, src_txt, tgt_txt = b_data |
| 70 | sent_data = {"src": indexed_tokens, "segs": segments_ids} |
| 71 | |
| 72 | src = torch.tensor(_pad([sent_data['src']], 0, pad_size)).to(device) |
| 73 | segs = torch.tensor(_pad([sent_data['segs']], 0, pad_size)).to(device) |
| 74 | mask = torch.logical_not(src == 0).to(device) |
| 75 | sentence_vector = model(src, segs, mask) |
| 76 | |
| 77 | return sentence_vector |
| 78 | |
| 79 | |
| 80 | def add_vector_in_origin_file(path, vector_dict, save_path): |
nothing calls this directly
no test coverage detected