bpe-en-gl_emb.yaml

save_data: run
## Where the vocab(s) will be written
src_vocab: run/bpe.vocab.src
tgt_vocab: run/bpe.vocab.tgt
overwrite: True

# Corpus opts:
data:
    europarl:
        path_src: ../DGTcorpora_tokenized/en_gl/europarl/partitions/en_train.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/europarl/partitions/gl_train.txt
        transforms: [bpe, filtertoolong]
        weight: 120
    opensub:
        path_src: ../DGTcorpora_tokenized/en_gl/opensub/partitions/en_train.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/opensub/partitions/gl_train.txt
        transforms: [bpe, filtertoolong]
        weight: 152
    opus:
        path_src: ../DGTcorpora_tokenized/en_gl/opus/partitions/en_train.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/opus/partitions/gl_train.txt
        transforms: [bpe, filtertoolong]
        weight: 160
    ted2020:
        path_src: ../DGTcorpora_tokenized/en_gl/ted2020/partitions/en_train.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/ted2020/partitions/gl_train.txt
        transforms: [bpe, filtertoolong]
        weight: 10
    corgaback:
        path_src: ../DGTcorpora_tokenized/en_gl/corgaback/partitions/en_train.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/corgaback/partitions/gl_train.txt
        transforms: [bpe, filtertoolong]
        weight: 15
    ccmatrix:
        path_src: ../DGTcorpora_tokenized/en_gl/ccmatrix/en_tok_dbo.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/ccmatrix/gl_tok_dbo.txt
        transforms: [bpe, filtertoolong]
        weight: 380 ##75 ## 25000000/13000000 = 2; 760/2 = 380 * 5 = 1900 (380/5=75)
    wikimatrix:
        path_src: ../DGTcorpora_tokenized/en_gl/wikimatrix/en.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/wikimatrix/gl.txt
        transforms: [bpe, filtertoolong]
        weight: 70 #25000000/450000 = 55 ; 760/55 = 14 ; 14 * 5 = 70
    cluvi:
        path_src: ../DGTcorpora_tokenized/en_gl/cluvi/en.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/cluvi/gl.txt
        transforms: [bpe, filtertoolong]
        weight: 70 #25000000/295000 = 84 ; 760/84 = 9 ; 9 * 10 = 90
        #wikimedia:
        # path_src: ../DGTcorpora_tokenized/en_gl/wikimedia/en.txt
        #path_tgt: ../DGTcorpora_tokenized/en_gl/wikimedia/gl.txt
        #transforms: [bpe, filtertoolong]
        #weight: 4
        # xlent:
        #path_src: ../DGTcorpora_tokenized/en_gl/xlent/en.txt
        #path_tgt: ../DGTcorpora_tokenized/en_gl/xlent/gl.txt
        #transforms: [bpe, filtertoolong]
        #weight: 50 #25000000/1600000=15; 760/15=50
        #linux:
        #path_src: ../DGTcorpora_tokenized/en_gl/linux/en.txt
        #path_tgt: ../DGTcorpora_tokenized/en_gl/linux/gl.txt
        #transforms: [bpe, filtertoolong]
        #weight: 20 #25000000/150000=166; 760/166=5 * 5 = 20
    valid:
        path_src: ../DGTcorpora_tokenized/en_gl/partitions/all-en_valid.txt
        path_tgt: ../DGTcorpora_tokenized/en_gl/partitions/all-gl_valid.txt
        transforms: [bpe, filtertoolong]

### Transform related opts:
#### Subword
src_subword_model: ./bpe/en.code
tgt_subword_model: ./bpe/gl.code
src_subword_vocab: ./run/bpe.vocab.src
tgt_subword_vocab: ./run/bpe.vocab.tgt
#src_subword_model: ../sentencepiece/en-gl/en.sp.model
#tgt_subword_model: ../sentencepiece/en-gl/gl.sp.model
src_subword_type: bpe
tgt_subord_type: bpe

src_subword_nbest: 1
src_subword_alpha: 0.0
tgt_subword_nbest: 1
tgt_subword_alpha: 0.0
#### Filter
src_seq_length: 150
tgt_seq_length: 150

# silently ignore empty lines in the data
skip_empty_level: silent

##embeddings
src_embeddings: ../embeddings/en.emb.txt
tgt_embeddings: ../embeddings/gl.emb.txt

## supported types: GloVe, word2vec
embeddings_type: "word2vec"

# word_vec_size need to match with the pretrained embeddings dimensions
word_vec_size: 300



# General opts
save_model: run/model
keep_checkpoint: 50
save_checkpoint_steps: 10000
average_decay: 0.0005
seed: 1234
report_every: 1000
train_steps: 200000
valid_steps: 10000

# Batching
queue_size: 10000
bucket_size: 32768
world_size: 1
gpu_ranks: [0]
batch_type: "tokens"
batch_size: 8192
#batch_size: 4096
valid_batch_size: 64
batch_size_multiple: 1
max_generator_batches: 2
accum_count: [4]
accum_steps: [0]

# Optimization
model_dtype: "fp16"
optim: "adam"
learning_rate: 2
warmup_steps: 8000
decay_method: "noam"
adam_beta2: 0.998
max_grad_norm: 0
label_smoothing: 0.1
param_init: 0
param_init_glorot: true
normalization: "tokens"

# Model
encoder_type: transformer
decoder_type: transformer
position_encoding: true
enc_layers: 6
dec_layers: 6
heads: 8
rnn_size: 512
word_vec_size: 512
transformer_ff: 2048
dropout_steps: [0]
dropout: [0.1]
attention_dropout: [0.1]
share_decoder_embeddings: true
share_embeddings: false