actableai.third_parties.spanABSA.absa.run_base.bert_load_state_dict(model, state_dict)¶actableai.third_parties.spanABSA.absa.run_base.copy_optimizer_params_to_model(named_params_model, named_params_optimizer)¶Utility function for optimize_on_cpu and 16-bits training. Copy the parameters optimized on CPU/RAM back to the model on GPU
actableai.third_parties.spanABSA.absa.run_base.post_process_loss(args, n_gpu, loss)¶actableai.third_parties.spanABSA.absa.run_base.prepare_optimizer(args, model, num_train_steps)¶actableai.third_parties.spanABSA.absa.run_base.set_optimizer_params_grad(named_params_optimizer, named_params_model, test_nan=False)¶Utility function for optimize_on_cpu and 16-bits training. Copy the gradient of the GPU parameters to the CPU/RAMM copy of the model
Run BERT on SemEval.
actableai.third_parties.spanABSA.absa.run_cls_span.eval_absa(all_examples, all_features, all_results, do_lower_case, verbose_logging, logger)¶actableai.third_parties.spanABSA.absa.run_cls_span.evaluate(args, model, device, eval_examples, eval_features, eval_dataloader, logger, write_pred=False)¶actableai.third_parties.spanABSA.absa.run_cls_span.main()¶actableai.third_parties.spanABSA.absa.run_cls_span.metric_max_over_ground_truths(metric_fn, term, polarity, gold_terms, gold_polarities)¶actableai.third_parties.spanABSA.absa.run_cls_span.pipeline_eval_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_cls_span.read_eval_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_cls_span.read_train_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_cls_span.run_train_epoch(args, global_step, model, param_optimizer, train_dataloader, eval_examples, eval_features, eval_dataloader, optimizer, n_gpu, device, logger, log_path, save_path, save_checkpoints_steps, start_save_steps, best_f1)¶Run BERT on SemEval.
actableai.third_parties.spanABSA.absa.run_extract_span.eval_aspect_extract(all_examples, all_features, all_results, do_lower_case, verbose_logging, logger)¶actableai.third_parties.spanABSA.absa.run_extract_span.evaluate(args, model, device, eval_examples, eval_features, eval_dataloader, logger, write_pred=False, do_pipeline=False)¶actableai.third_parties.spanABSA.absa.run_extract_span.main()¶actableai.third_parties.spanABSA.absa.run_extract_span.read_eval_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_extract_span.read_train_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_extract_span.run_train_epoch(args, global_step, model, param_optimizer, train_dataloader, eval_examples, eval_features, eval_dataloader, optimizer, n_gpu, device, logger, log_path, save_path, save_checkpoints_steps, start_save_steps, best_f1)¶Run BERT on SemEval.
actableai.third_parties.spanABSA.absa.run_joint_span.evaluate(args, model, device, eval_examples, eval_features, eval_dataloader, logger, write_pred=False)¶actableai.third_parties.spanABSA.absa.run_joint_span.main()¶actableai.third_parties.spanABSA.absa.run_joint_span.read_eval_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_joint_span.read_train_data(args, tokenizer, logger)¶actableai.third_parties.spanABSA.absa.run_joint_span.run_train_epoch(args, global_step, model, param_optimizer, train_examples, train_features, train_dataloader, eval_examples, eval_features, eval_dataloader, optimizer, n_gpu, device, logger, log_path, save_path, save_checkpoints_steps, start_save_steps, best_f1)¶actableai.third_parties.spanABSA.absa.utils.InputFeatures(unique_id, example_index, tokens, token_to_orig_map, input_ids, input_mask, segment_ids, start_positions=None, end_positions=None, start_indexes=None, end_indexes=None, bio_labels=None, polarity_positions=None, polarity_labels=None, label_masks=None)¶Bases: object
A single set of features of data.
actableai.third_parties.spanABSA.absa.utils.RawBIOClsResult(unique_id, start_indexes, end_indexes, bio_pred, span_masks)¶Bases: tuple
bio_pred¶Alias for field number 3
end_indexes¶Alias for field number 2
span_masks¶Alias for field number 4
start_indexes¶Alias for field number 1
unique_id¶Alias for field number 0
actableai.third_parties.spanABSA.absa.utils.RawBIOResult(unique_id, bio_pred)¶Bases: tuple
bio_pred¶Alias for field number 1
unique_id¶Alias for field number 0
actableai.third_parties.spanABSA.absa.utils.RawFinalResult(unique_id, start_indexes, end_indexes, cls_pred, span_masks)¶Bases: tuple
cls_pred¶Alias for field number 3
end_indexes¶Alias for field number 2
span_masks¶Alias for field number 4
start_indexes¶Alias for field number 1
unique_id¶Alias for field number 0
actableai.third_parties.spanABSA.absa.utils.RawSpanCollapsedResult(unique_id, neu_start_logits, neu_end_logits, pos_start_logits, pos_end_logits, neg_start_logits, neg_end_logits)¶Bases: tuple
neg_end_logits¶Alias for field number 6
neg_start_logits¶Alias for field number 5
neu_end_logits¶Alias for field number 2
neu_start_logits¶Alias for field number 1
pos_end_logits¶Alias for field number 4
pos_start_logits¶Alias for field number 3
unique_id¶Alias for field number 0
actableai.third_parties.spanABSA.absa.utils.RawSpanResult(unique_id, start_logits, end_logits)¶Bases: tuple
end_logits¶Alias for field number 2
start_logits¶Alias for field number 1
unique_id¶Alias for field number 0
actableai.third_parties.spanABSA.absa.utils.SemEvalExample(example_id, sent_tokens, term_texts=None, start_positions=None, end_positions=None, polarities=None)¶Bases: object
actableai.third_parties.spanABSA.absa.utils.convert_absa_data(dataset, verbose_logging=False)¶actableai.third_parties.spanABSA.absa.utils.convert_examples_to_features(examples, tokenizer, max_seq_length, verbose_logging=False, logger=None)¶actableai.third_parties.spanABSA.absa.utils.pos2term(words, starts, ends)¶actableai.third_parties.spanABSA.absa.utils.read_absa_data(path)¶read data from the specified path :param path: path of dataset :return:
actableai.third_parties.spanABSA.absa.utils.span_annotate_candidates(all_examples, batch_features, batch_results, filter_type, is_training, use_heuristics, use_nms, logit_threshold, n_best_size, max_answer_length, do_lower_case, verbose_logging, logger)¶Annotate top-k candidate answers into features.
actableai.third_parties.spanABSA.absa.utils.ts2polarity(words, ts_tag_sequence, starts, ends)¶actableai.third_parties.spanABSA.absa.utils.ts2start_end(ts_tag_sequence)¶actableai.third_parties.spanABSA.absa.utils.wrapped_get_final_text(example, feature, start_index, end_index, do_lower_case, verbose_logging, logger)¶