Python misc.utils.LanguageModelCriterion() Examples

The following are 7 code examples of misc.utils.LanguageModelCriterion(). 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 misc.utils , or try the search function .
Example #1
Source File: loss_wrapper.py    From GoogleConceptualCaptioning with MIT License 6 votes vote down vote up 
def __init__(self, model, opt):
        super(LossWrapper, self).__init__()
        self.opt = opt
        self.model = model
        if opt.label_smoothing > 0:
            self.crit = utils.LabelSmoothing(smoothing=opt.label_smoothing)
        else:
            self.crit = utils.LanguageModelCriterion()
        self.rl_crit = utils.RewardCriterion()
        self.struc_crit = utils.StructureLosses(opt)


        if opt.vse_model != 'None':
            self.vse = VSEFCModel(opt)
            for p in self.vse.parameters():
                p.requires_grad = False
            self.retrieval_reward_weight = opt.retrieval_reward_weight # 

            self.vse.load_state_dict({k[4:]:v for k,v in torch.load(opt.initialize_retrieval).items() if 'vse.' in k})
        self.retrieval_reward_weight = 0
Example #2
Source File: eval.py    From video-caption.pytorch with MIT License 6 votes vote down vote up 
def main(opt):
    dataset = VideoDataset(opt, "test")
    opt["vocab_size"] = dataset.get_vocab_size()
    opt["seq_length"] = dataset.max_len
    if opt["model"] == 'S2VTModel':
        model = S2VTModel(opt["vocab_size"], opt["max_len"], opt["dim_hidden"], opt["dim_word"],
                          rnn_dropout_p=opt["rnn_dropout_p"]).cuda()
    elif opt["model"] == "S2VTAttModel":
        encoder = EncoderRNN(opt["dim_vid"], opt["dim_hidden"], bidirectional=opt["bidirectional"],
                             input_dropout_p=opt["input_dropout_p"], rnn_dropout_p=opt["rnn_dropout_p"])
        decoder = DecoderRNN(opt["vocab_size"], opt["max_len"], opt["dim_hidden"], opt["dim_word"],
                             input_dropout_p=opt["input_dropout_p"],
                             rnn_dropout_p=opt["rnn_dropout_p"], bidirectional=opt["bidirectional"])
        model = S2VTAttModel(encoder, decoder).cuda()
    #model = nn.DataParallel(model)
    # Setup the model
    model.load_state_dict(torch.load(opt["saved_model"]))
    crit = utils.LanguageModelCriterion()

    test(model, crit, dataset, dataset.get_vocab(), opt)
Example #3
Source File: loss_wrapper.py    From AAT with MIT License 5 votes vote down vote up 
def __init__(self, model, opt):
        super(LossWrapper, self).__init__()
        self.opt = opt
        self.model = model
        if opt.label_smoothing > 0:
            self.crit = utils.LabelSmoothing(smoothing=opt.label_smoothing)
        else:
            self.crit = utils.LanguageModelCriterion()
        self.rl_crit = utils.RewardCriterion()
Example #4
Source File: loss_wrapper.py    From AoANet with MIT License 5 votes vote down vote up 
def __init__(self, model, opt):
        super(LossWrapper, self).__init__()
        self.opt = opt
        self.model = model
        if opt.label_smoothing > 0:
            self.crit = utils.LabelSmoothing(smoothing=opt.label_smoothing)
        else:
            self.crit = utils.LanguageModelCriterion()
        self.rl_crit = utils.RewardCriterion()
Example #5
Source File: eval.py    From video-caption.pytorch with MIT License 5 votes vote down vote up 
def main(opt):
    dataset = VideoDataset(opt, "test")
    opt["vocab_size"] = dataset.get_vocab_size()
    opt["seq_length"] = dataset.max_len
    if opt['beam_size'] != 1:
        assert opt["batch_size"] == 1
    if opt["model"] == 'S2VTModel':
        model = S2VTModel(opt["vocab_size"], opt["max_len"], opt["dim_hidden"], opt["dim_word"], opt['dim_vid'],
                          n_layers=opt['num_layers'],
                          rnn_cell=opt['rnn_type'],
                          bidirectional=opt["bidirectional"],
                          rnn_dropout_p=opt["rnn_dropout_p"]).cuda()
    elif opt["model"] == "S2VTAttModel":
        encoder = EncoderRNN(opt["dim_vid"], opt["dim_hidden"],
                             n_layers=opt['num_layers'],
                             rnn_cell=opt['rnn_type'], bidirectional=opt["bidirectional"],
                             input_dropout_p=opt["input_dropout_p"], rnn_dropout_p=opt["rnn_dropout_p"])
        decoder = DecoderRNN(opt["vocab_size"], opt["max_len"], opt["dim_hidden"], opt["dim_word"],
                             n_layers=opt['num_layers'],
                             rnn_cell=opt['rnn_type'], input_dropout_p=opt["input_dropout_p"],
                             rnn_dropout_p=opt["rnn_dropout_p"], bidirectional=opt["bidirectional"])
        model = S2VTAttModel(encoder, decoder).cuda()
    model = nn.DataParallel(model)
    # Setup the model
    model.load_state_dict(torch.load(opt["saved_model"]))
    crit = utils.LanguageModelCriterion()

    test(model, crit, dataset, dataset.get_vocab(), opt)
Example #6
Source File: train.py    From video-caption.pytorch with MIT License 4 votes vote down vote up 
def main(opt):
    dataset = VideoDataset(opt, 'train')
    dataloader = DataLoader(dataset, batch_size=opt["batch_size"], shuffle=True)
    opt["vocab_size"] = dataset.get_vocab_size()
    if opt["model"] == 'S2VTModel':
        model = S2VTModel(
            opt["vocab_size"],
            opt["max_len"],
            opt["dim_hidden"],
            opt["dim_word"],
            opt['dim_vid'],
            rnn_cell=opt['rnn_type'],
            n_layers=opt['num_layers'],
            bidirectional=opt["bidirectional"],
            rnn_dropout_p=opt["rnn_dropout_p"]).cuda()
    elif opt["model"] == "S2VTAttModel":
        encoder = EncoderRNN(
            opt["dim_vid"],
            opt["dim_hidden"],
            n_layers=opt['num_layers'],
            bidirectional=opt["bidirectional"],
            input_dropout_p=opt["input_dropout_p"],
            rnn_cell=opt['rnn_type'],
            rnn_dropout_p=opt["rnn_dropout_p"])
        decoder = DecoderRNN(
            opt["vocab_size"],
            opt["max_len"],
            opt["dim_hidden"],
            opt["dim_word"],
            n_layers=opt['num_layers'],
            input_dropout_p=opt["input_dropout_p"],
            rnn_cell=opt['rnn_type'],
            rnn_dropout_p=opt["rnn_dropout_p"],
            bidirectional=opt["bidirectional"])
        model = S2VTAttModel(encoder, decoder).cuda()
    crit = utils.LanguageModelCriterion()
    rl_crit = utils.RewardCriterion()
    optimizer = optim.Adam(
        model.parameters(),
        lr=opt["learning_rate"],
        weight_decay=opt["weight_decay"])
    exp_lr_scheduler = optim.lr_scheduler.StepLR(
        optimizer,
        step_size=opt["learning_rate_decay_every"],
        gamma=opt["learning_rate_decay_rate"])

    train(dataloader, model, crit, optimizer, exp_lr_scheduler, opt, rl_crit)
Example #7
Source File: train.py    From video-caption.pytorch with MIT License 4 votes vote down vote up 
def main(opt):
    dataset = VideoDataset(opt, 'train')
    dataloader = DataLoader(dataset, batch_size=opt["batch_size"], shuffle=True)
    opt["vocab_size"] = dataset.get_vocab_size()
    if opt["model"] == 'S2VTModel':
        model = S2VTModel(
            opt["vocab_size"],
            opt["max_len"],
            opt["dim_hidden"],
            opt["dim_word"],
            opt['dim_vid'],
            rnn_cell=opt['rnn_type'],
            n_layers=opt['num_layers'],
            rnn_dropout_p=opt["rnn_dropout_p"])
    elif opt["model"] == "S2VTAttModel":
        encoder = EncoderRNN(
            opt["dim_vid"],
            opt["dim_hidden"],
            bidirectional=opt["bidirectional"],
            input_dropout_p=opt["input_dropout_p"],
            rnn_cell=opt['rnn_type'],
            rnn_dropout_p=opt["rnn_dropout_p"])
        decoder = DecoderRNN(
            opt["vocab_size"],
            opt["max_len"],
            opt["dim_hidden"],
            opt["dim_word"],
            input_dropout_p=opt["input_dropout_p"],
            rnn_cell=opt['rnn_type'],
            rnn_dropout_p=opt["rnn_dropout_p"],
            bidirectional=opt["bidirectional"])
        model = S2VTAttModel(encoder, decoder)
    model = model.cuda()
    crit = utils.LanguageModelCriterion()
    rl_crit = utils.RewardCriterion()
    optimizer = optim.Adam(
        model.parameters(),
        lr=opt["learning_rate"],
        weight_decay=opt["weight_decay"])
    exp_lr_scheduler = optim.lr_scheduler.StepLR(
        optimizer,
        step_size=opt["learning_rate_decay_every"],
        gamma=opt["learning_rate_decay_rate"])

    train(dataloader, model, crit, optimizer, exp_lr_scheduler, opt, rl_crit)