KeyError: "word '•' not in vocabulary"

ikuinen / CMIN_moment_retrieval

Cross-Modal Interaction Networks for Query-Based Moment Retrieval in Videos

86 stars 20 forks source link

Closed SCZwangxiao closed 4 years ago

SCZwangxiao commented 4 years ago

I am using the GloVe model glove.42B.300d.bin and the training processing is successful. However in model.eval( ) function an error occured:

I 've located the abnormal data in the evaluation dataset:

id: v_ah3tGziTbds
sentence: "The logo \"International Sailing Academy Laz Cruz \u2022 Mexico\" appears on screen."

It seems that word2vec cannot encode word '\u2022'. Could you tell me how did you deal with this?

SCZwangxiao commented 4 years ago

Problem solved when using GloVe model 'glove.840B.300d.bin'