eric-wallace / adversarial-mt Goto Github PK

Hi I was trying out to setup the github repo codes. I followed the entire setup process but I encountered this error:
I have installed the requirements.txt file before this and run through the process.
Is it possible to provide the full requirements.txt file with the versions of the dependencies inside? The current one only has the package but not the versions. Appreciate your time! Thank you!

RuntimeError: result type Float can't be cast to the desired output type Long

Full error log as follows:
Traceback (most recent call last):
File "targeted_flips.py", line 160, in
main()
File "targeted_flips.py", line 32, in main
targeted_flips(samples, args, trainer, generator, embedding_weight, bpe)
File "targeted_flips.py", line 46, in targeted_flips
translations = trainer.task.inference_step(generator, [trainer.get_model()], samples)
File "/data/home/boonghim/adversarial-mt-master/fairseq/tasks/fairseq_task.py", line 265, in inference_step
return generator.generate(models, sample, prefix_tokens=prefix_tokens)
File "/data/home/boonghim/.conda/envs/attacking/lib/python3.6/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context
return func(*args, **kwargs)
File "/data/home/boonghim/adversarial-mt-master/fairseq/sequence_generator.py", line 113, in generate
return self._generate(model, sample, **kwargs)
File "/data/home/boonghim/.conda/envs/attacking/lib/python3.6/site-packages/torch/autograd/grad_mode.py", line 26, in decorate_context
return func(*args, **kwargs)
File "/data/home/boonghim/adversarial-mt-master/fairseq/sequence_generator.py", line 379, in _generate
scores.view(bsz, beam_size, -1)[:, :, :step],
File "/data/home/boonghim/adversarial-mt-master/fairseq/search.py", line 83, in step
out=self.beams_buf)

Hi Eric,

Just here to ask if the project can run automatically on multiple sentences. Like, we have an input of tens of sentences, and the project can do the flip on all sentences without inputting sentences one by one?

Thank you for your kind answer in advance.