cbaziotis / ntua-slp-semeval2018 Goto Github PK

Deep-learning models of NTUA-SLP team submitted in SemEval 2018 tasks 1, 2 and 3.

License: MIT License

Python 100.00%

deep-learning lstm sentiment-analysis affect-recognition emotion-recognition irony-detection semeval attention attention-mechanism

ntua-slp-semeval2018's Issues

code takes longer to pre process trian data

No cache file for TASK3_A_word_train ...
PreProcessing dataset TASK3_A_word_train...: 2%|▏ | 63/3834 [00:00<00:40, 93.01it/s]

It keeps waiting at 2%.

Even when I try to load using GPU or only CPU. gold set was very fast and took 3 minutes to process

predict_models only shows two results for each dataset

The json file in the out directory contains only two results for each dataset.

Error with load_embeddings.py "can't fix"

Kindly, help me to solve the problem with this error which I receive when doing the steps to run
python -m model.pretraining.sentiment2017

File "/utils/load_embeddings.py", line 93, in load_word_vectors
embeddings = numpy.array(embeddings, dtype=numpy.float32)
ValueError: setting an array element with a sequence.

Thanks

Which average Jaccard I should use?

In Task 1 5 multi-label class.
Which one Jaccard I should use? The source code doesn't have the definition.
micro ? macro? weighted?

How to train and evaluate the TASK2_A?

Hello

I've run the TASK2_A with several embeddings file via its configuration (model/params.py) through command below:

python3 model.task2.train.py

And I've got the following result:

Running on:cuda
loading word embeddings...
Didn't find embeddings cache file /content/gdrive/My Drive/app/sentiment/glove.twitter.27B.200d.txt
Indexing file /content/gdrive/My Drive/app/sentiment/glove.twitter.27B.200d.txt ...
{200}
Found 1193516 word vectors.
Reading twitter_2018 - 1grams ...
Reading twitter_2018 - 2grams ...
Reading twitter_2018 - 1grams ...
Building word-level datasets...
Loading TASK2_A_train from cache!
Total words: 8286649, Total unks:918613 (11.09%)
Unique words: 160897, Unique unks:81783 (50.83%)
Labels statistics:
{0: '21.72%', 1: '10.47%', 2: '10.28%', 3: '5.53%', 4: '4.98%', 5: '4.71%', 6: '4.31%', 7: '3.70%', 8: '3.44%', 9: '3.26%', 10: '3.25%', 11: '3.10%', 12: '2.80%', 13: '2.62%', 14: '2.73%', 15: '2.71%', 16: '2.64%', 17: '2.59%', 18: '2.68%', 19: '2.49%'}

Loading TASK2_A_trial from cache!
Total words: 814679, Total unks:77013 (9.45%)
Unique words: 40332, Unique unks:10207 (25.31%)
Labels statistics:
{0: '21.52%', 1: '10.56%', 2: '10.48%', 3: '5.77%', 4: '5.03%', 5: '4.63%', 6: '4.10%', 7: '3.79%', 8: '3.59%', 9: '3.34%', 10: '3.09%', 11: '3.06%', 12: '2.92%', 13: '2.69%', 14: '2.75%', 15: '2.50%', 16: '2.61%', 17: '2.56%', 18: '2.57%', 19: '2.43%'}

Loading TASK2_A_gold from cache!
Total words: 895048, Total unks:103962 (11.62%)
Unique words: 42546, Unique unks:11668 (27.42%)
Labels statistics:
{0: '21.60%', 1: '9.66%', 2: '9.07%', 3: '5.21%', 4: '7.43%', 5: '3.23%', 6: '3.99%', 7: '5.50%', 8: '3.10%', 9: '2.35%', 10: '2.86%', 11: '3.90%', 12: '2.53%', 13: '2.23%', 14: '2.61%', 15: '2.49%', 16: '2.31%', 17: '3.09%', 18: '4.83%', 19: '2.02%'}

Where is the F1-score result?
What did I miss?

Could someone tell me which version of python and cuda did you use please? Thanks very much!

When I run python model/task1/train.py , it comes this error:

RuntimeError: CUDNN_STATUS_EXECUTION_FAILED

I guess it's because my Python or CUDA versions aren't right, can someone tell me which version of python and cuda did you use?
Thanks in advance!

How to trained the ntu_310 that has 10 affective embedding?

Error when training the model for Task2A

I am trying to train the neural network model for Task2A (i.e., emoji prediction in English), from SemEval2018. However, when I run the file neural_experiment.py, located in model/task2/, I get the following error message:

File "./model/task2/neural_experiment.py", line 52, in <module>
    model_training(trainer, model_config["epochs"])

  File "../..\utils\train.py", line 387, in model_training
    trainer.train()

  File "../..\logger\training.py", line 520, in train
    self.train_loader(self.loaders["train"])

  File "../..\logger\training.py", line 488, in train_loader
    sample_batched)

  File "../..\utils\train.py", line 143, in pipeline
    outputs, attentions = nn_model(inputs, lengths)

  File "D:\WinPython\python-3.7.2.amd64\lib\site-packages\torch\nn\modules\module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)

  File "../..\modules\nn\models.py", line 203, in forward
    representations, attentions = self.feature_extractor(x, lengths)

  File "D:\WinPython\python-3.7.2.amd64\lib\site-packages\torch\nn\modules\module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)

  File "../..\modules\nn\models.py", line 140, in forward
    embeddings = self.embedding(x)

  File "D:\WinPython\python-3.7.2.amd64\lib\site-packages\torch\nn\modules\module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)

  File "../..\modules\nn\modules.py", line 148, in forward
    embeddings = self.embedding(x)

  File "D:\WinPython\python-3.7.2.amd64\lib\site-packages\torch\nn\modules\module.py", line 489, in __call__
    result = self.forward(*input, **kwargs)

  File "D:\WinPython\python-3.7.2.amd64\lib\site-packages\torch\nn\modules\sparse.py", line 118, in forward
    self.norm_type, self.scale_grad_by_freq, self.sparse)

  File "D:\WinPython\python-3.7.2.amd64\lib\site-packages\torch\nn\functional.py", line 1454, in embedding
    return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)

RuntimeError: Expected tensor for argument #1 'indices' to have scalar type Long; but got torch.IntTensor instead (while checking arguments for embedding)

Do you have any idea what I should do next?

IndexError: tuple index out of range

Hi, I get this error when running sentiment_2017.py
The embeddings loaded are those from ntua_twitter_300.txt

Could someone help me with that?
Here the error message:

Traceback (most recent call last):
File "/home/ignacioalvarez/PycharmProjects/ntua-slp-semeval2018/model/pretraining/sentiment2017.py", line 39, in
monitor="val", label_transformer=transformer)
File "/home/ignacioalvarez/PycharmProjects/ntua-slp-semeval2018/utils/train.py", line 338, in define_trainer
**_config)
File "/home/ignacioalvarez/PycharmProjects/ntua-slp-semeval2018/modules/nn/models.py", line 184, in init
**kwargs)
File "/home/ignacioalvarez/PycharmProjects/ntua-slp-semeval2018/modules/nn/models.py", line 85, in init
embedding_dim=embeddings.shape[1],
IndexError: tuple index out of range

Pretraining Getting Stuck

I am running the pretraining code the way you suggested but it has been stuck at this point for 2 hours now. Is this supposed to take this long?

neilpaul77@NeilRig77:~/Downloads/ntua-slp-semeval2018$ python sentiment2017.py 
/home/neilpaul77/anaconda3/lib/python3.6/site-packages/sklearn/externals/joblib/externals/cloudpickle/cloudpickle.py:47: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
  import imp
Running on:cuda
loading word embeddings...
Loaded word embeddings from cache.
Reading twitter_2018 - 1grams ...
Reading twitter_2018 - 2grams ...
Reading twitter_2018 - 1grams ...
Building word-level datasets...
Loading SEMEVAL_2017_word_train from cache!
Total words: 1435889, Total unks:9700 (0.68%)
Unique words: 45397, Unique unks:3602 (7.93%)
Labels statistics:
{'negative': '18.91%', 'neutral': '45.47%', 'positive': '35.62%'}

Loading SEMEVAL_2017_word_val from cache!
Total words: 75465, Total unks:521 (0.69%)
Unique words: 9191, Unique unks:198 (2.15%)
Labels statistics:
{'negative': '18.91%', 'neutral': '45.46%', 'positive': '35.63%'}

Initializing Embedding layer with pre-trained weights!
ModelWrapper(
  (feature_extractor): FeatureExtractor(
    (embedding): Embed(
      (embedding): Embedding(804871, 310)
      (dropout): Dropout(p=0.1)
      (noise): GaussianNoise (mean=0.0, stddev=0.2)
    )
    (encoder): RNNEncoder(
      (rnn): LSTM(310, 150, num_layers=2, batch_first=True, dropout=0.3, bidirectional=True)
      (drop_rnn): Dropout(p=0.3)
    )
    (attention): SelfAttention(
      (attention): Sequential(
        (0): Linear(in_features=300, out_features=300, bias=True)
        (1): Tanh()
        (2): Dropout(p=0.3)
        (3): Linear(in_features=300, out_features=1, bias=True)
        (4): Tanh()
        (5): Dropout(p=0.3)
      )
      (softmax): Softmax()
    )
  )
  (linear): Linear(in_features=300, out_features=3, bias=True)
)

prediction of Task 3

hi, I want to reproduce the result of Task 3.
And now I can run the base_experiment of task 3.
But I don't know how to train and test.
Can you give me some instruction?
Thanks a lot.

problems with ekphrasis

Hi,

I'm trying to run your code on GPU by running the following command:
python ./model/pretraining/sentiment2017.py

but the following errors occur. Can you help me with that ?

Running on:cuda Traceback (most recent call last): File "./model/pretraining/sentiment2017.py", line 17, in <module> from utils.train import define_trainer, model_training File "/root/ntua-slp-semeval2018/utils/train.py", line 21, in <module> from modules.nn.dataloading import WordDataset, CharDataset File "/root/ntua-slp-semeval2018/modules/nn/dataloading.py", line 7, in <module> from ekphrasis.classes.preprocessor import TextPreProcessor File "/usr/local/lib/python2.7/dist-packages/ekphrasis/classes/preprocessor.py", line 6, in <module> from ekphrasis.classes.exmanager import ExManager File "/usr/local/lib/python2.7/dist-packages/ekphrasis/classes/exmanager.py", line 19 {print(k.lower(), ":", self.expressions[k]) ^ SyntaxError: invalid syntax

models are not saved

Hi,

models of task 1 and 3 are not saved. The output is:


Traceback (most recent call last):
  File "/usr/local/lib/python3.5/dist-packages/visdom/__init__.py", line 388, in _send
    data=json.dumps(msg),
  File "/usr/lib/python3/dist-packages/requests/api.py", line 107, in post
    return request('post', url, data=data, json=json, **kwargs)
  File "/usr/lib/python3/dist-packages/requests/api.py", line 53, in request
    return session.request(method=method, url=url, **kwargs)
  File "/usr/lib/python3/dist-packages/requests/sessions.py", line 468, in request
    resp = self.send(prep, **send_kwargs)
  File "/usr/lib/python3/dist-packages/requests/sessions.py", line 576, in send
    r = adapter.send(request, **kwargs)
  File "/usr/lib/python3/dist-packages/requests/adapters.py", line 437, in send
    raise ConnectionError(e, request=request)
requests.exceptions.ConnectionError: HTTPConnectionPool(host='localhost', port=8097): Max retries exceeded with url: /update (Caused by NewConnectionError('<requests.packages.urllib3.connection.HTTPConnection object at 0x7f8b80177b00>: Failed to establish a new connection: [Errno 111] Connection refused',))

patience left:-1, best(0.768163003566172)
Early stopping...

code examples of training and a sample for prediction

Hi, many thanks for your contribution.

Is there any example of how to input a simple text massage into the model and get a prediction?

Testing working setup does not work

After following all prerequisites, starting the visdom server and trying to run python model/pretraining/sentiment2017.py I get the following error

Traceback (most recent call last):
  File "model/pretraining/sentiment2017.py", line 9, in <module>
    from config import DATA_DIR
ModuleNotFoundError: No module named 'config'

which makes complete sense, as there is no config.py file in that folder. I've tried moving the config.py file found in the root directory to that folder, and after running it again, it gives the following error

Running on:cpu
Traceback (most recent call last):
  File "model/pretraining/sentiment2017.py", line 10, in <module>
    from dataloaders.rest import load_data_from_dir
ModuleNotFoundError: No module named 'dataloaders'

I've then also moved sentiment2017.py to the root folder, and then run it from there with python sentiment2017.py, which sort of works. It starts up, takes about 5 minutes to get to ~4%, and then just hangs there as far as I can tell.

PreProcessing dataset SEMEVAL_2017_word_train...:   4%| | 2240/58761 [05:10<2:10:32,  7.22it/s]

Any idea what is happening? I am using Python 3.6 (and also had to install some packages that aren't included in the requirements.txt), but when using Python 2.7 I get a different error. Also not sure why the instructions in the README don't work for me and I have to copy and move files around either.

Full Documentation of Source Code

Thank you for your contribution. Are you still planning to add the full documentation of the source code?

In particular, I am interested in understanding the difference between the scripts for task 1.

cbaziotis / ntua-slp-semeval2018 Goto Github PK

ntua-slp-semeval2018's Issues

code takes longer to pre process trian data

predict_models only shows two results for each dataset

Error with load_embeddings.py "can't fix"

Which average Jaccard I should use?

How to train and evaluate the TASK2_A?

Could someone tell me which version of python and cuda did you use please? Thanks very much!

How to trained the ntu_310 that has 10 affective embedding?

Error when training the model for Task2A

IndexError: tuple index out of range

Pretraining Getting Stuck

prediction of Task 3

problems with ekphrasis

models are not saved

code examples of training and a sample for prediction

Testing working setup does not work

Full Documentation of Source Code

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent