xinnuoxu/mmi_anti_pytorch

A pytorch implementation for the MMI-anti model

mmi_anti_pytorch's Issues

hello

handsome boy, excuse me!
when I have been studing chatbot, I have a question that I don't understand . because of I thank you are enthusiastic,so I want to ask you for this question as follow:

eng_vocab_size = len(eng_tokenizer.word_index) + 1

this code means vocabulary size equals the length of word-index adding 1, I dont't understand why it need to add 1. I think it is necessary to add 1. please you help me , thank you !

Error in Training

Traceback (most recent call last):
File "translate.py", line 157, in
main()
File "translate.py", line 119, in main
batch_data = translator.translate_batch(batch, data)
File "/home/thea/cikm/mmi_anti_pytorch/onmt/translate/Translator.py", line 205, in translate_batch
ret = self._from_beam(beam)
File "/home/thea/cikm/mmi_anti_pytorch/onmt/translate/Translator.py", line 234, in _from_beam
best_scores, best_scores_id = self.tt.FloatTensor(b.ended_sentences_scores).topk(n_best, 0, True, True)
RuntimeError: invalid argument 3: dim not in range at /pytorch/aten/src/TH/generic/THTensorMath.c:3436

Hi XinnouXu, I caught an error in training 'python translate.py -model modelname' and really need some help here.
Thanks in advance!

version of pytorch, torchtext

Which version of pytorch and torchtext do you use ?
I got the following message when trying to perform translate.py.
(I use pytorch 0.3.0 and torchtext 0.2.1. )

TypeError: torch.index_select received an invalid combination of arguments - got (torch.cuda.FloatTensor, int, torch.LongTensor), but expected (torch.cuda.FloatTensor source, int dim, torch.cuda.LongTensor index)

where does the mmi loss function locate?

I just see two losscompute in train.py
`def make_loss_compute(model, tgt_vocab, opt):

if opt.copy_attn:
    compute = onmt.modules.CopyGeneratorLossCompute(
        model.generator, tgt_vocab, opt.copy_attn_force)
else:
    compute = onmt.Loss.NMTLossCompute(
        model.generator, tgt_vocab,
        label_smoothing=opt.label_smoothing)

if use_gpu(opt):
    compute.cuda()

return compute`

TypeError: __init__() got an unexpected keyword argument 'tensor_type'

I use pytorch-cpu 1.0.1， torchtext0.3.1, I am new for pytorch. How can i solve this problem?

`Traceback (most recent call last):
File "/home/yangjingying/PycharmProjects/mmi_anti_pytorch-master/train.py", line 403, in
Loading train dataset from data/dialogue.train.1.pt, number of examples: 100
main()
File "/home/yangjingying/PycharmProjects/mmi_anti_pytorch-master/train.py", line 385, in main
fields = load_fields(first_dataset, data_type, checkpoint)
File "/home/yangjingying/PycharmProjects/mmi_anti_pytorch-master/train.py", line 306, in load_fields
torch.load(opt.data + '.vocab.pt'), data_type)
File "/home/yangjingying/PycharmProjects/mmi_anti_pytorch-master/onmt/io/IO.py", line 57, in load_fields_from_vocab
fields = get_fields(data_type, n_src_features, n_tgt_features)
File "/home/yangjingying/PycharmProjects/mmi_anti_pytorch-master/onmt/io/IO.py", line 43, in get_fields
return TextDataset.get_fields(n_src_features, n_tgt_features)
File "/home/yangjingying/PycharmProjects/mmi_anti_pytorch-master/onmt/io/TextDataset.py", line 210, in get_fields
postprocessing=make_src, sequential=False)

TypeError: init() got an unexpected keyword argument 'tensor_type'
`

代码跑不通

Where is “lm/model.pt”？

I can't find the model.pt.

indentation in omnt/translate/Translator.py

There are many problems for indentation for the file Translator.py . Please fix it.