initialbug / biset Goto Github PK
View Code? Open in Web Editor NEWBiSET: Bi-directional Selective Encoding with Template for Abstractive Summarization (ACL 2019)
License: MIT License
BiSET: Bi-directional Selective Encoding with Template for Abstractive Summarization (ACL 2019)
License: MIT License
I didn't execute the “Retrieve” module, I think it took a little bit of time. You provided retrived and reranked data in "Notice". I downloaded the data from "Google Disk", which contains x.title.ext, x.template.txt, x.title.txt, test.template.rerank-x. I want to have a look at the file "x.t.plate.index", and I want to execute the second module "FastRerank", could you please provide x.samples.index.json, x_scores.json, x.template.index?
Hi,
Is it possible to either provide the generated outputs on the test set or any pretrained model checkpoint that I can directly use to decode the summaries ? I only need the summaries of the test set.
Thanks!
Excuse me, I am wondering what is the dataset for retrieve?
Is it same as the training set of source articles, English Gigaword?
I am really not familiar with Information Retrieve, please forgive me if this is a stupid question. Thank you!
In config.py
, on line 26-28
it seems like the preprocessing
step need article.txt
, title.txt
, template
, samples.index.json
and _score.json
to be prepared in config.py
to run the whole process
But after doing retrieve, I only got train/test/dev.sample.index
other than the original article and title file
So how can I get all the other data I need such as sample.index.json
and _score.json
?
I have read this article, but I did not find any introduction on how to generate a template!could you please tell me how to generate a template?
when I was trying to train my model with -copy_attn
(copy attention function)
It occurs some errors in both condition of whether using gpu or not
But I'm thinking the template setting does not conflict with this copy_attn function (also the coverage_attn
works well )
so my command line looks like python3 train.py -data path/to/data -copy_attn
and the error message looks like (with gpu)
Traceback (most recent call last):
File "train.py", line 42, in <module>
main(opt)
File "train.py", line 28, in main
single_main(opt)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/train_single.py", line 133, in main
opt.valid_steps)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/trainer.py", line 172, in train
report_stats)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/trainer.py", line 296, in _gradient_accumulation
trunc_size, self.shard_size, normalization)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/utils/loss.py", line 145, in sharded_compute_loss
loss, stats = self._compute_loss(batch, **shard)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/modules/copy_generator.py", line 193, in _compute_loss
batch, self.tgt_vocab, batch.dataset.src_vocabs)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/inputters/text_dataset.py", line 119, in collapse_copy_scores
print('index: %s'%index)
File "/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/tensor.py", line 71, in __repr__
return torch._tensor_str._str(self)
File "/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/_tensor_str.py", line 286, in _str
tensor_str = _tensor_str(self, indent)
File "/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/_tensor_str.py", line 201, in _tensor_str
formatter = _Formatter(get_summarized_data(self) if summarize else self)
File "/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/_tensor_str.py", line 83, in __init__
value_str = '{}'.format(value)
File "/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/tensor.py", line 386, in __format__
return self.item().__format__(format_spec)
RuntimeError: CUDA error: device-side assert triggered
and the error message without using gpu is like
/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torchtext/data/field.py:359: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
var = torch.tensor(arr, dtype=self.dtype, device=device)
/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/nn/functional.py:1386: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
/home/yoshow/venv/pytorch11/lib/python3.5/site-packages/torch/nn/functional.py:1374: UserWarning: nn.functional.tanh is deprecated. Use torch.tanh instead.
warnings.warn("nn.functional.tanh is deprecated. Use torch.tanh instead.")
Traceback (most recent call last):
File "train.py", line 42, in <module>
main(opt)
File "train.py", line 28, in main
single_main(opt)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/train_single.py", line 133, in main
opt.valid_steps)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/trainer.py", line 172, in train
report_stats)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/trainer.py", line 296, in _gradient_accumulation
trunc_size, self.shard_size, normalization)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/utils/loss.py", line 145, in sharded_compute_loss
loss, stats = self._compute_loss(batch, **shard)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/modules/copy_generator.py", line 189, in _compute_loss
loss = self.criterion(scores, align, target)
File "/media/yoshow/HDD/BiSET/Bi-selective Encoding/onmt/modules/copy_generator.py", line 125, in __call__
out = scores.gather(1, align.view(-1, 1) + self.offset).view(-1)
RuntimeError: Invalid index in gather at /pytorch/aten/src/TH/generic/THTensorEvenMoreMath.cpp:459
I trained the model successfully and ran the command $ python3 translate.py -model model_step_85000.pt -src path_to_data/test.article.txt -template path_to_data/test.template.rerank-5
for translation
but it occurred the following error:
Traceback (most recent call last):
File "translate.py", line 36, in <module>
main(opt)
File "translate.py", line 25, in main
attn_debug=opt.attn_debug)
File "/home/yoshow/Summarization/BiSET/Bi-selective Encoding/onmt/translate/translator.py", line 198, in translate
use_filter_pred=self.use_filter_pred)
File "/home/yoshow/Summarization/BiSET/Bi-selective Encoding/onmt/inputters/inputter.py", line 248, in build_dataset
use_filter_pred=use_filter_pred)
File "/home/yoshow/Summarization/BiSET/Bi-selective Encoding/onmt/inputters/text_dataset.py", line 66, in __init__
ex, examples_iter = self._peek(examples_iter)
File "/home/yoshow/Summarization/BiSET/Bi-selective Encoding/onmt/inputters/dataset_base.py", line 107, in _peek
first = next(seq)
File "/home/yoshow/Summarization/BiSET/Bi-selective Encoding/onmt/inputters/text_dataset.py", line 303, in _dynamic_dict
template = example["template"]
KeyError: 'template'
not sure where does this error come from
Hello, I am not sure that I set all parameters right. Can you provide some parameters? Thanks.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.