java_small_nofeat_share.yml

data:
    corpus_1:
        path_src: ../synpos/data/java-small-processed/train_src.txt
        path_tgt: ../synpos/data/java-seq2seq-data/data.TargetType.seq.train.target.txt
        transforms: [filterfeats, onmt_tokenize, inferfeats] #, filtertoolong]
        weight: 1
    valid:
        path_src: ../synpos/data/java-small-processed/val_src.txt
        path_tgt: ../synpos/data/java-seq2seq-data/data.TargetType.seq.val.target.txt
        transforms: [filterfeats, onmt_tokenize, inferfeats]

src_onmttok_kwargs: "{'mode': 'space'}"
tgt_onmttok_kwargs: "{'mode': 'space'}"

# # Vocab opts
src_vocab: java_small_transformer_sum/data_shared/data.vocab.src
tgt_vocab: java_small_transformer_sum/data_shared/data.vocab.src

save_data: java_small_data
overwrite: False

#save_model: java_small_transformer_nofeat_copy_share/model
save_checkpoint_steps: 1000
keep_checkpoint: 20
seed: 3435
train_steps: 500000
valid_steps: 1000
valid_batch_size: 8192
warmup_steps: 8000
report_every: 100
src_vocab_size: 2000
share_embeddings: True
share_vocab: True
copy_attn: True
# Uri:
#early_stopping: 10 
#early_stopping_criteria: accuracy

decoder_type: transformer
encoder_type: transformer
word_vec_size: 512
rnn_size: 512
layers: 6
transformer_ff: 2048
heads: 8

accum_count: 8
optim: adam
adam_beta1: 0.9
adam_beta2: 0.998
decay_method: noam
learning_rate: 2.0
max_grad_norm: 0.0

batch_size: 4096
batch_type: tokens
normalization: tokens
dropout: 0.1
label_smoothing: 0.1

max_generator_batches: 2

param_init: 0.0
param_init_glorot: 'true'
position_encoding: 'true'

world_size: 2
gpu_ranks:
- 0
- 1