Closed ryanachi closed 2 years ago
Looks like there was another error in the stack trace that was hard to spot:
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/flair/models/sequence_tagger_model.py", line 232, in __init__
self.to(flair.device)
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/module.py", line 673, in to
return self._apply(convert)
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/module.py", line 387, in _apply
module._apply(fn)
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/module.py", line 387, in _apply
module._apply(fn)
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/module.py", line 387, in _apply
module._apply(fn)
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/flair/models/language_model.py", line 428, in _apply
child_module._apply(fn)
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 186, in _apply
self.flatten_parameters()
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 170, in flatten_parameters
if self.proj_size > 0:
File "/home/ryanchi/miniconda3/envs/glare/lib/python3.7/site-packages/torch/nn/modules/module.py", line 948, in __getattr__
type(self).__name__, name))
AttributeError: 'LSTM' object has no attribute 'proj_size'
This actually might be a problem with the flair version. @ryanachi can you run the following command and paste the output? pip show flair
(the most recent FLAIR version is 0.8: https://github.com/flairNLP/flair)
Describe the bug When running textattack's CLARE recipe, it fails to generate more than one result, exiting due to the following error:
TypeError: 'NoneType' object is not callable
(please see the paste2 for full details)To Reproduce Steps to reproduce the behavior:
textattack attack --recipe clare --model bert-base-uncased-ag-news --num-examples 20
Expected behavior Given an example dataset, the CLARE recipe should (drawing on a RoBERTa model) create adversarial examples through infilling, merging, and replacing tokens.
Screenshots or Traceback https://paste2.org/gtc0V94W
System Information (please complete the following information):
transformers==4.5.1
)Thank you for your help!