Closed ioana-blue closed 5 years ago
First, thanks for sharing the code, saved me quite a bit of sweat :)
Second, there might be a typo in the instructions to build the vocabulary files. I'm (wild)guessing that the following lines should refer to the target file (instead of inputs):
ognn-build-vocab --with_sequence_tokens \ --save_vocab /data/naturallanguage/cnn_dailymail/output.vocab \ /data/naturallanguage/cnn_dailymail/split/train/inputs.jsonl.gz
should read
ognn-build-vocab --with_sequence_tokens \ --save_vocab /data/naturallanguage/cnn_dailymail/output.vocab \ /data/naturallanguage/cnn_dailymail/split/train/targets.jsonl.gz
if I understand correctly what this vocab file is meant for.
Thanks for this. Should be fixed
First, thanks for sharing the code, saved me quite a bit of sweat :)
Second, there might be a typo in the instructions to build the vocabulary files. I'm (wild)guessing that the following lines should refer to the target file (instead of inputs):
should read
if I understand correctly what this vocab file is meant for.