keyerror

[zkt12]

Why there's a keyerror when I run predict on dev_v2.1.json? Here is the traceback: "Augmenting with pre-trained embeddings... Augmenting with random char embeddings... Traceback (most recent call last): File "scripts/predict.py", line 189, in main() File "scripts/predict.py", line 182, in main toks = ' '.join(id_to_token[tok] for tok in toks) File "scripts/predict.py", line 182, in toks = ' '.join(id_to_token[tok] for tok in toks) KeyError: tensor(507946)"

[spacemanidol]

Hey,

The predict.py and train.py files are works in progress and do not work yet. To start your baseline please refer to the duet.py file. While not fully productionalized it is a fully functional baseline.

[zkt12]

Sorry, I didn't make it clear in my last question. I run predict in the Q+A task, not in the ranking task. Why did the keyerror happen?

[spacemanidol]

Hey I dont know what may be going on for you. I've been testing and I created a clean ubuntu box, downloaded the data and repo, followed the instructions for training(trained for 1 epoch) and then ran prediction no problem both with no custom embeddings and using glove. I think this is likley related to your embeddings.

erasmus@spacemanidol:~/MSMARCOV2/Q+A/Baseline$ python3 scripts/predict.py exp/ ../data/dev_v2.1.json prediction.json --cuda=True --word_rep ../data/glove.840B.300d.txt /usr/local/lib/python3.5/dist-packages/h5py/init.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type. from ._conv import register_converters as _register_converters Loading Model... Augmenting with pre-trained embeddings... Augmenting with random char embeddings... erasmus@spacemanidol:~/MSMARCOV2/Q+A/Baseline$