Failed to download the saved_embd.pt

[JinmingZhao]

Hi， When I download the data.zip and will prompt the error. Now I have the multinli_0.9 and snli datasets, but don't have the saved_embed.pt. So, I want to know the if saved_embed.pt. is the word2vec vectors ? If it is， I will extracted the features from the fasttext-based word2vec models as mentioned in InferSent paper.

Thanks

[KhenAharon]

I created the missing file "saved_embd.pt". You can download it from my drive: https://drive.google.com/open?id=1vDJfjEUGnK-S3gZ5sv6Q_PDeJ0P3tR6q

The way I created it is by using torchtext to download pretrained GloVe vectors (840B, 300D). I used the newest version of torchtext to create it (not the old version mentioned in the Author's README since the API is different). Then you should have an attribute that is called "vectors" in the downloaded object. Use torch to save those vectors in a .pt format (this format is a convention for PyTorch saved files - "pt" files). When you load it in the res_encoder you should use the old torchtext version as the authors mentioned.