870e17a6857365bec3035bfece2e77b525288b01,exp.py,,train,#Any#,50

Before Change

                eval_results = evaluation.evaluate(dev_set.examples, model, args,
                                                   verbose=True, eval_top_pred_only=args.eval_top_pred_only)
                dev_score = eval_results["bleu"] if args.lang == "conala" else eval_results["accuracy"]
                if args.lang == "conala":
                    print("[Epoch %d] evaluate details: %s took %ds" % (
                                        epoch, eval_results, time.time() - eval_start), file=sys.stderr)
                else:
                    print("[Epoch %d] code generation accuracy=%.5f took %ds" % (
                                        epoch, dev_score, time.time() - eval_start), file=sys.stderr)
                is_better = history_dev_scores == [] or dev_score > max(history_dev_scores)
                history_dev_scores.append(dev_score)
        else:
            is_better = True

After Change

    parser_cls = get_parser_class(args.lang)
    model = parser_cls(args, vocab, transition_system)
    model.train()
    evaluator = Registrable.by_name(args.evaluator)(transition_system)
    if args.cuda: model.cuda()

    optimizer_cls = eval("torch.optim.%s" % args.optimizer)  // FIXME: this is evil!
    optimizer = optimizer_cls(model.parameters(), lr=args.lr)

Instances