diff --git a/config/code2seq-java-med.yaml b/config/code2seq-java-med.yaml new file mode 100644 index 0000000..82fc17d --- /dev/null +++ b/config/code2seq-java-med.yaml @@ -0,0 +1,56 @@ +data_folder: ../data/code2seq/java-med + +checkpoint: null + +seed: 7 +log_offline: false +# Training in notebooks (e.g. Google Colab) may crash with too small value +progress_bar_refresh_rate: 1 +print_config: true + +data: + url: https://s3.eu-west-1.amazonaws.com/datasets.ml.labs.aws.intellij.net/java-paths-methods/java-med.tar.gz + num_workers: 4 + + # Each token appears at least 5 times + max_labels: 13935 + max_label_parts: 7 + # Each token appears at least 5 times + max_tokens: 62063 + max_token_parts: 5 + path_length: 9 + + max_context: 200 + random_context: true + + batch_size: 512 + test_batch_size: 768 + +model: + # Encoder + embedding_size: 128 + encoder_dropout: 0.25 + encoder_rnn_size: 128 + use_bi_rnn: true + rnn_num_layers: 1 + + # Decoder + decoder_size: 320 + decoder_num_layers: 1 + rnn_dropout: 0.5 + +optimizer: + optimizer: "Momentum" + nesterov: true + lr: 0.01 + weight_decay: 0 + decay_gamma: 0.95 + +train: + n_epochs: 10 + patience: 10 + clip_norm: 5 + teacher_forcing: 1.0 + val_every_epoch: 1 + save_every_epoch: 1 + log_every_n_steps: 10 \ No newline at end of file