diff options
Diffstat (limited to 'tests/training/data-weighting')
8 files changed, 12 insertions, 12 deletions
diff --git a/tests/training/data-weighting/test_maxi_batches_with_sentence_weights.sh b/tests/training/data-weighting/test_maxi_batches_with_sentence_weights.sh index 9abd743..0c39b5e 100644 --- a/tests/training/data-weighting/test_maxi_batches_with_sentence_weights.sh +++ b/tests/training/data-weighting/test_maxi_batches_with_sentence_weights.sh @@ -11,7 +11,7 @@ test -e vocab.de.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-e test -e vocab.en.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-en/corpus.bpe.en > vocab.en.yml $MRT_MARIAN/build/marian \ - --seed 3333 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 3333 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m maxibatch/model.npz -t train.1k.{de,en} -v vocab.{de,en}.yml \ --log maxibatch.log --disp-freq 10 --after-batches 100 --mini-batch 16 --cost-type ce-sum \ --data-weighting train.1k.inc.txt --data-weighting-type sentence diff --git a/tests/training/data-weighting/test_maxi_batches_with_word_weights.sh b/tests/training/data-weighting/test_maxi_batches_with_word_weights.sh index 4982d19..2228796 100644 --- a/tests/training/data-weighting/test_maxi_batches_with_word_weights.sh +++ b/tests/training/data-weighting/test_maxi_batches_with_word_weights.sh @@ -11,7 +11,7 @@ test -e vocab.de.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-e test -e vocab.en.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-en/corpus.bpe.en > vocab.en.yml $MRT_MARIAN/build/marian \ - --seed 6666 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 6666 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m word_maxibatch/model.npz -t train.1k.{de,en} -v vocab.{de,en}.yml \ --log word_maxibatch.log --disp-freq 10 --after-batches 100 --mini-batch 16 \ --data-weighting train.1k.wordinc.txt --data-weighting-type word diff --git a/tests/training/data-weighting/test_sentence_weighting_sqlite.sh b/tests/training/data-weighting/test_sentence_weighting_sqlite.sh index 93f9bee..7323c2d 100644 --- a/tests/training/data-weighting/test_sentence_weighting_sqlite.sh +++ b/tests/training/data-weighting/test_sentence_weighting_sqlite.sh @@ -8,7 +8,7 @@ rm -rf sqlite sqlite.log mkdir -p sqlite $MRT_MARIAN/build/marian \ - --seed 1111 --no-shuffle --maxi-batch 1 --maxi-batch-sort none --max-length 100 --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 1111 --no-shuffle --maxi-batch 1 --maxi-batch-sort none --max-length 100 --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m sqlite/model.npz -t train.1k.{de,en} -v vocab.{de,en}.yml \ --log sqlite.log --disp-freq 1 --after-batches 100 --mini-batch 1 \ --data-weighting train.1k.weights.txt --data-weighting-type sentence --sqlite sqlite/corpus.sqlite3 diff --git a/tests/training/data-weighting/test_sentence_weighting_with_ones.sh b/tests/training/data-weighting/test_sentence_weighting_with_ones.sh index 7063c80..caeba16 100644 --- a/tests/training/data-weighting/test_sentence_weighting_with_ones.sh +++ b/tests/training/data-weighting/test_sentence_weighting_with_ones.sh @@ -11,7 +11,7 @@ test -e vocab.de.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-e test -e vocab.en.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-en/corpus.bpe.en > vocab.en.yml $MRT_MARIAN/build/marian \ - --seed 2222 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 2222 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m noweights/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log noweights.log --disp-freq 5 -e 2 @@ -22,7 +22,7 @@ cat noweights.log | $MRT_TOOLS/strip-timestamps.sh | grep "Ep\. " | sed -r 's/ T cat $MRT_DATA/europarl.de-en/toy.bpe.en | sed -r 's/.*/1/g' > ones.weights.txt $MRT_MARIAN/build/marian \ - --seed 2222 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 2222 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m ones/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log ones.log --disp-freq 5 -e 2 \ --data-weighting ones.weights.txt diff --git a/tests/training/data-weighting/test_sentence_weights_x3.sh b/tests/training/data-weighting/test_sentence_weights_x3.sh index 893e5ab..11cec28 100644 --- a/tests/training/data-weighting/test_sentence_weights_x3.sh +++ b/tests/training/data-weighting/test_sentence_weights_x3.sh @@ -11,7 +11,7 @@ test -e vocab.de.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-e test -e vocab.en.yml || $MRT_MARIAN/build/marian-vocab < $MRT_DATA/europarl.de-en/corpus.bpe.en > vocab.en.yml $MRT_MARIAN/build/marian \ - --seed 2222 --no-shuffle --maxi-batch 1 --maxi-batch-sort none --max-length 100 --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 2222 --no-shuffle --maxi-batch 1 --maxi-batch-sort none --max-length 100 --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m x3copied/model.npz -t train.x3.{de,en} -v vocab.{de,en}.yml \ --log x3copied.log --disp-freq 1 --after-batches 100 --mini-batch 4 --cost-type ce-sum @@ -20,7 +20,7 @@ test -e x3copied.log cat x3copied.log | grep 'Cost ' | sed -r 's/.*Cost (.*) : Time.*/\1/' > x3copied.out $MRT_MARIAN/build/marian \ - --seed 2222 --no-shuffle --maxi-batch 1 --maxi-batch-sort none --max-length 100 --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 2222 --no-shuffle --maxi-batch 1 --maxi-batch-sort none --max-length 100 --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m x3weights/model.npz -t train.1k.{de,en} -v vocab.{de,en}.yml \ --log x3weights.log --disp-freq 1 --after-batches 100 --mini-batch 2 --cost-type ce-sum \ --data-weighting train.1k.weights.txt --data-weighting-type sentence diff --git a/tests/training/data-weighting/test_word_weighting_sqlite.sh b/tests/training/data-weighting/test_word_weighting_sqlite.sh index 2421f5e..33c276e 100644 --- a/tests/training/data-weighting/test_word_weighting_sqlite.sh +++ b/tests/training/data-weighting/test_word_weighting_sqlite.sh @@ -10,7 +10,7 @@ mkdir -p sqlite_word cat $MRT_DATA/europarl.de-en/toy.bpe.en | sed -r 's/[^ ]+/2/g' > sqlite_word.weights.txt $MRT_MARIAN/build/marian \ - --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m sqlite_word/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log sqlite_word.log --disp-freq 5 -e 2 --mini-batch-fit -w 500 \ --data-weighting sqlite_word.weights.txt --data-weighting-type word --sqlite sqlite_word/corpus.sqlite3 diff --git a/tests/training/data-weighting/test_word_weighting_with_ones.sh b/tests/training/data-weighting/test_word_weighting_with_ones.sh index 082032e..8ba31a8 100644 --- a/tests/training/data-weighting/test_word_weighting_with_ones.sh +++ b/tests/training/data-weighting/test_word_weighting_with_ones.sh @@ -8,7 +8,7 @@ rm -rf word_noweights* word_ones* mkdir -p word_noweights word_ones $MRT_MARIAN/build/marian \ - --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m word_noweights/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log word_noweights.log --disp-freq 5 -e 2 @@ -19,7 +19,7 @@ cat word_noweights.log | $MRT_TOOLS/strip-timestamps.sh | grep "Ep\. " | sed -r cat $MRT_DATA/europarl.de-en/toy.bpe.en | sed -r 's/[^ ]+/1/g' > word_ones.weights.txt $MRT_MARIAN/build/marian \ - --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m word_ones/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log word_ones.log --disp-freq 5 -e 2 \ --data-weighting word_ones.weights.txt --data-weighting-type word diff --git a/tests/training/data-weighting/test_word_weighting_with_twos.sh b/tests/training/data-weighting/test_word_weighting_with_twos.sh index 0bfcc4c..4706cae 100644 --- a/tests/training/data-weighting/test_word_weighting_with_twos.sh +++ b/tests/training/data-weighting/test_word_weighting_with_twos.sh @@ -10,7 +10,7 @@ mkdir -p word_twos cat $MRT_DATA/europarl.de-en/toy.bpe.en | sed -r 's/[^ ]+/2/g' > word_twos.weights.txt $MRT_MARIAN/build/marian \ - --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m word_twos/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log word_twos.log --disp-freq 5 -e 2 \ --data-weighting word_twos.weights.txt --data-weighting-type word @@ -28,7 +28,7 @@ echo "data-weighting: word_twos.weights.txt" > word_twos.config.yml echo "data-weighting-type: word" >> word_twos.config.yml $MRT_MARIAN/build/marian \ - --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 -o sgd \ + --seed 1111 --no-shuffle --dim-emb 128 --dim-rnn 256 --optimizer sgd \ -m word_twos_cfg/model.npz -t $MRT_DATA/europarl.de-en/toy.bpe.{de,en} -v vocab.{de,en}.yml \ --log word_twos_cfg.log --disp-freq 5 -e 2 \ -c word_twos.config.yml |