structured-neural-summarization/models/java-model.sh at master · mloncode/structured-neural-summarization · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
python train_and_eval.py --train_source_file data/sourcecode/java/split/train/inputs.jsonl.gz \
                         --train_target_file data/sourcecode/java/split/train/summaries.jsonl.gz \
                         --valid_source_file data/sourcecode/java/split/valid/inputs.jsonl.gz \
                         --valid_target_file data/sourcecode/java/split/valid/summaries.jsonl.gz \
                         --node_vocab_file data/sourcecode/java/node.vocab \
                         --edge_vocab_file data/sourcecode/java/edge.vocab \
                         --target_vocab_file data/sourcecode/java/output.vocab \
                         --train_steps 50000 \
                         --optimizer momentum \
                         --learning_rate 0.1 \
                         --lr_decay_rate 0.95 \
                         --lr_decay_steps 5000 \
                         --copy_attention  \
                         --model_name java_small_momentum_0.1_256_0.0  \
                         --checkpoint_dir java_small_momentum_0.1_256_0.0 \
                         --rnn_hidden_size 256 \
                         --rnn_hidden_dropout 0.0 \
                         --node_features_dropout 0.0 \
                         --validation_interval 5000 \
                         --embeddings_dropout 0.0 \
                         --case_sensitive  \
                         --attend_all_nodes