rescore.py 文件源码

def main(models, source_file, nbest_file, saveto, b=80,
         normalize=False, verbose=False, alignweights=False):

    # load model model_options
    options = []
    for model in args.models:
        try:
            with open('%s.json' % model, 'rb') as f:
                options.append(json.load(f))
        except:
            with open('%s.pkl' % model, 'rb') as f:
                options.append(pkl.load(f))
        #hacks for using old models with missing options
        if not 'dropout_embedding' in options[-1]:
            options[-1]['dropout_embedding'] = 0
        if not 'dropout_hidden' in options[-1]:
            options[-1]['dropout_hidden'] = 0
        if not 'dropout_source' in options[-1]:
            options[-1]['dropout_source'] = 0
        if not 'dropout_target' in options[-1]:
            options[-1]['dropout_target'] = 0

    dictionary, dictionary_target = options[0]['dictionaries']

    # load source dictionary and invert
    word_dict = load_dict(dictionary)
    word_idict = dict()
    for kk, vv in word_dict.iteritems():
        word_idict[vv] = kk
    word_idict[0] = '<eos>'
    word_idict[1] = 'UNK'

    # load target dictionary and invert
    word_dict_trg = load_dict(dictionary_target)
    word_idict_trg = dict()
    for kk, vv in word_dict_trg.iteritems():
        word_idict_trg[vv] = kk
    word_idict_trg[0] = '<eos>'
    word_idict_trg[1] = 'UNK'

    rescore_model(source_file, nbest_file, saveto, models, options, b, normalize, verbose, alignweights)