aswalin / squad Goto Github PK

I downloaded the pickle file using the command
curl -Lo encoder/infersent1.pkl https://dl.fbaipublicfiles.com/senteval/infersent/infersent1.pkl
and import it using
infersent = torch.load('infersent1.pkl', map_location=lambda storage, loc: storage) infersent.set_glove_path("glove.840B.300d.txt")
the error which i got -
"AttributeError: 'collections.OrderedDict' object has no attribute 'set_glove_path'"

I also updated the file models.py in working directory with InferSent class but still not able to load it using infersent.set_w2v_path.
What is correct way to overcome this?

I tried to reproduce your create_emb.ipynb but I have issues with

infersent = torch.load('InferSent/encoder/infersent.allnli.pickle', map_location=lambda storage, loc: storage)

I tried to look on InterSent/encoder folder here but nothing appears to be named infersent.allnli.pickle. Maybe you're referencing an old version ? Maybe I'm wrong because I neither see the infersent.pickle they are referencing.

Getting UnpicklingError: invalid load key, '<'. while loading the pre-trained model infersent.allnli.pickle

Hi,

May I know your running time for getting the embedding matrix by InferSent? It was running over 4 hours on my computer and still not finished, not sure if anything went wrong.. Thanks.

I tried downloading infersent1.pickle and interfece2.pickle but it is getting error. I think we must provide infersent.allnli.pickle this file only.but that is downloading corrupt file.