assertion error about laser HOT 5 CLOSED

I suppose that the embeddings were not calculated correctly, and consequently, you get a size mismatch with the class target file.
This is probably related to #9 , i.e. the fastBPE command was not in the path.
Please pull the latest source and try again (removing spurious files)

from laser.

@hoschwenk I have the latest source and the modification mentioned in #9 is also done. But the error is still there.

the link provided to download the XNLI 1.0 and XNLI-MT 1.0 seems to be broken. can this be the issue?

from laser.

I tied this solution, it solved a part of the problem, but this one still remain :

Training the classifier (see embed/xnli.log)
Traceback (most recent call last):
File "/home/trotignon/projects/LASER/source/nli.py", line 228, in
dim=args.dim, bsize=args.bsize, shuffle=True)
File "/home/trotignon/projects/LASER/source/nli.py", line 57, in LoadDataNLI
D = data_utils.TensorDataset(torch.from_numpy(nli), torch.from_numpy(lbl))
File "/home/trotignon/anaconda3/lib/python3.6/site-packages/torch/utils/data/dataset.py", line 36, in init
assert all(tensors[0].size(0) == tensor.size(0) for tensor in tensors)
AssertionError

Even with the last version,
Thank you

from laser.

Indeed, it looks like that the XNLI link is broken or was moved elsewhere.
Sadly, I can't help you with this. If there is no data, the tool won't work :-)
Please have a look at the size of the created files - I guess that there is an error somewhere.

from laser.

Thank you very much, will look at it!

from laser.