Train:

• make-vocab:

python -m allennlp.run train experiments/baseline_vocab.json  -s tmp/baseline/ --include-package abstractive_summarization

• replace:

cat experiments/baseline.json > tmp/baseline/config.json

• train-recover:

python -m allennlp.run train experiments/baseline.json  -r -s tmp/baseline/ --include-package abstractive_summarization

• vis:

cd tmp/baseline/log/train
tensorboard --logdir=./

predcit

python -m allennlp.run predict tmp/baseline/model.tar.gz tmp/test.jsonl --batch-size 32 --cuda-device 0 --include-package abstractive_summarization --predictor abstract-generator

model.tar.gz

tar cvf model.tar *.*
    vocabulary/
    vocabulary/non_padded_namespaces.txt
    vocabulary/tokens.txt
    weights.json
    weights.th
gzip .model.tar