Create bpe-gl-es_emb.yaml

bpe-gl-es_emb.yaml

@@ -0,0 +1,155 @@
+
+save_data: run
+## Where the vocab(s) will be written
+src_vocab: run/vocab/es-gl/bpe.vocab.src
+tgt_vocab: run/vocab/es-gl/bpe.vocab.tgt
+overwrite: True
+
+# Corpus opts:
+data:
+    europarl:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/europarl/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/europarl_translit/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 120 #60 #120
+    opensub:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/opensub/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/opensub_translit/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 180 #900 #180
+    dgt:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/dgt/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/dgt_translit/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 18 #9 #18
+    cluvi:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/cluvi/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/cluvi/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 40 # 4 #40
+    opensub-es-gl:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/opensub-es-gl/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/opensub-es-gl/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 25 # 5 #25 #25
+    ted2020:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/ted2020/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/ted2020/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 10 # 1 #10 #10
+    corgaback:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/corgaback/partitions/es_train.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/corgaback/partitions/gl_train.txt
+        transforms: [bpe, filtertoolong]
+        weight: 13 # 66  #14 #13
+    ccmatrix:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/ccmatrix/es.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/ccmatrix/gl.txt
+        transforms: [bpe, filtertoolong]
+        weight: 180 ##como opensub, tamanho semelhante
+    resto:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/resto/es.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/resto/gl.txt
+        transforms: [bpe, filtertoolong]
+        weight: 120 ##como europarl, tamanho semelhante
+    opensub_2018:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/opensub_2018/es.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/opensub_2018/gl.txt
+        transforms: [bpe, filtertoolong]
+        weight: 25 #igual que opensub_es-gl
+
+        
+    valid:
+        path_tgt: ../DGTcorpora_tokenized/es_gz/partitions/all-es_valid.txt
+        path_src: ../DGTcorpora_tokenized/es_gz/partitions_translit/all-gl_valid.txt
+        transforms: [bpe, filtertoolong]
+
+### Transform related opts:
+#### Subword
+tgt_subword_model: ./bpe/es.code
+src_subword_model: ./bpe/gl.code
+tgt_subword_vocab: ./run/vocab/es-gl/bpe.vocab.src
+src_subword_vocab: ./run/vocab/es-gl/bpe.vocab.tgt
+#tgt_subword_model: ../sentencepiece/en-gl/en.sp.model
+#src_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
+
+##embeddings
+tgt_embeddings: ../embeddings/es.emb.txt
+src_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
+
+
+#### Filter
+src_seq_length: 150
+tgt_seq_length: 150
+
+# silently ignore empty lines in the data
+skip_empty_level: silent
+
+
+
+# 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: 4096
+batch_size: 8192
+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
+#learning_rate: 0.00005
+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