Python opts.py() Examples

The following are 2 code examples of opts.py(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module opts , or try the search function .
Example #1
Source File: train.py    From reversible-rnn with MIT License 4 votes vote down vote up 
def evaluate(best_val_checkpoint_path):
    # python translate.py -src data/multi30k/test2016.en.atok -output pred.txt \
    #                     -replace_unk -tgt=data/multi30k/test2016.de.atok -report_bleu -gpu 2
    #                     -model saves/2018-02-09-enc:Rev-dec:Rev-et:RevGRU-dt:RevGRU-h:300-el:1-dl:1-em:300-atn:general-cxt:slice_emb-sl:20-ef1:0.875-ef2:0.875-df1:0.875-df2:0.875/best_checkpoint.pt

    base_dir = os.path.dirname(best_val_checkpoint_path)

    if '600' in best_val_checkpoint_path:
        test_output = subprocess.run(['python', 'translate.py', '-src', 'data/en-de/IWSLT16.TED.tst2014.en-de.en.tok.low',
                                      '-output', os.path.join(base_dir, 'test_pred.txt'), '-replace_unk', '-tgt', 'data/en-de/IWSLT16.TED.tst2014.en-de.de.tok.low',
                                      '-report_bleu', '-gpu', str(opt.gpuid[0]), '-model', best_val_checkpoint_path], stdout=subprocess.PIPE)

        test_output_string = test_output.stdout.decode('utf-8')
        print(test_output_string)

        # Also save the whole stdout string for reference
        with open(os.path.join(base_dir, 'test_stdout.txt'), 'w') as f:
            f.write('{}\n'.format(test_output_string))

        val_output = subprocess.run(['python', 'translate.py', '-src', 'data/en-de/IWSLT16.TED.tst2013.en-de.en.tok.low',
                                     '-output', os.path.join(base_dir, 'val_pred.txt'), '-replace_unk', '-tgt', 'data/en-de/IWSLT16.TED.tst2013.en-de.de.tok.low',
                                     '-report_bleu', '-gpu', str(opt.gpuid[0]), '-model', best_val_checkpoint_path], stdout=subprocess.PIPE)

        val_output_string = val_output.stdout.decode('utf-8')
        print(val_output_string)
    else:
        test_output = subprocess.run(['python', 'translate.py', '-src', 'data/multi30k/test2016.en.tok.low',
                                      '-output', os.path.join(base_dir, 'test_pred.txt'), '-replace_unk', '-tgt', 'data/multi30k/test2016.de.tok.low',
                                      '-report_bleu', '-gpu', str(opt.gpuid[0]), '-model', best_val_checkpoint_path], stdout=subprocess.PIPE)

        test_output_string = test_output.stdout.decode('utf-8')
        print(test_output_string)

        # Also save the whole stdout string for reference
        with open(os.path.join(base_dir, 'test_stdout.txt'), 'w') as f:
            f.write('{}\n'.format(test_output_string))

        val_output = subprocess.run(['python', 'translate.py', '-src', 'data/multi30k/val.en.tok.low',
                                     '-output', os.path.join(base_dir, 'val_pred.txt'), '-replace_unk', '-tgt', 'data/multi30k/val.de.tok.low',
                                     '-report_bleu', '-gpu', str(opt.gpuid[0]), '-model', best_val_checkpoint_path], stdout=subprocess.PIPE)

        val_output_string = val_output.stdout.decode('utf-8')
        print(val_output_string)

    # Also save the whole stdout string for reference
    with open(os.path.join(base_dir, 'val_stdout.txt'), 'w') as f:
        f.write('{}\n'.format(val_output_string))

    val_bleu = extract_bleu_score(val_output_string)
    test_bleu = extract_bleu_score(test_output_string)

    with open(os.path.join(base_dir, 'result.txt'), 'w') as f:
        f.write('{} {}\n'.format(val_bleu, test_bleu))

    print('Val BLEU: {} | Test BLEU: {}'.format(val_bleu, test_bleu))
Example #2
Source File: train.py    From QG-Net with MIT License 4 votes vote down vote up 
def main():

    # Load train and validate data.
    train_dataset = load_dataset("train")
    valid_dataset = load_dataset("valid")
    print(' * maximum batch size: %d' % opt.batch_size)

    # Load checkpoint if we resume from a previous training.
    if opt.train_from:
        print('Loading checkpoint from %s' % opt.train_from)
        checkpoint = torch.load(opt.train_from,
                                map_location=lambda storage, loc: storage)
        model_opt = checkpoint['opt']
        # I don't like reassigning attributes of opt: it's not clear.
        opt.start_epoch = checkpoint['epoch'] + 1
    else:
        checkpoint = None
        model_opt = opt

    # Load fields generated from preprocess phase.
    fields = load_fields(train_dataset, valid_dataset, checkpoint)

    # Report src/tgt features.
    collect_report_features(fields)

    # Build model.
    model = build_model(model_opt, opt, fields, checkpoint)
    tally_parameters(model)
    check_save_model_path()

    # Build optimizer.
    optim = build_optim(model, checkpoint)

    # load embeddings
    # NOTE you need to comment/uncomment the following section to use word embeddings!!!!!!
    # NOTE DO NOT USE THOSE WORD EMBED OPTIONS IN opts.py because they do not work!!!!!!
    fields['src'].vocab.load_vectors(wv_type='glove.42B', wv_dim=300)
    fields['tgt'].vocab.load_vectors(wv_type='glove.42B', wv_dim=300)
    model.encoder.embeddings.word_lut.weight.data.copy_(fields['src'].vocab.vectors.cuda())
    model.decoder.embeddings.word_lut.weight.data.copy_(fields['tgt'].vocab.vectors.cuda())
    model.encoder.embeddings.word_lut.weight.requires_grad = False
    model.decoder.embeddings.word_lut.weight.requires_grad = False

    # Do training.
    train_model(model, train_dataset, valid_dataset, fields, optim, model_opt)