Bi-directional Attention Flow (BiDAF) network is a multi-stage hierarchical process that represents context at different levels of granularity and uses a bi-directional attention flow mechanism to achieve a query-aware context representation without early summarization.
Hi ,
I am trying to train Bidaf on CNN data set but it fails.
Here are the steps performed.
Downloaded the CNN dataset
Ran the cnn_dm.prepro script. It generated json files properly
Run training using basic
python -m basic_cnn.cli --mode train --noload --debug
but it gives error that some training data file is not found. Looks like what is dumped in json file while pre-processing is different than what is expected by train script.
Hi , I am trying to train Bidaf on CNN data set but it fails. Here are the steps performed.
but it gives error that some training data file is not found. Looks like what is dumped in json file while pre-processing is different than what is expected by train script.
can you please help? Thanks