diff options
author | Hieu Hoang <hieuhoang@gmail.com> | 2014-12-30 16:15:26 +0300 |
---|---|---|
committer | Hieu Hoang <hieuhoang@gmail.com> | 2014-12-30 16:15:26 +0300 |
commit | 14cbf9bc2263485a5bef03fd61632a3fda41af6c (patch) | |
tree | 3256ea537f5a4be13ab325344f4b82a737da0b02 | |
parent | b82c31f020b2cd646a8874867a15c37126f2f1e7 (diff) | |
parent | 481a07dcc3338b9e2a6cbb2a9cf965bd68090806 (diff) |
Merge branch 'master' of github.com:moses-smt/mosesdecoder
-rw-r--r-- | scripts/ems/experiment.meta | 283 |
1 files changed, 52 insertions, 231 deletions
diff --git a/scripts/ems/experiment.meta b/scripts/ems/experiment.meta index d45dc849a..b059d5a81 100644 --- a/scripts/ems/experiment.meta +++ b/scripts/ems/experiment.meta @@ -20,9 +20,8 @@ clean out: clean-stem default-name: corpus/clean rerun-on-change: max-sentence-length $moses-script-dir/training/clean-corpus-n.perl - template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 $max-sentence-length OUT.lines-retained + template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 $max-sentence-length $working-dir/corpus/clean.lines-retained.VERSION error: there is a blank factor - error: is too long! at parse in: clean-stem out: parsed-stem @@ -36,7 +35,7 @@ post-parse-clean out: clean-parsed-stem default-name: corpus/parsed-clean pass-unless: input-parser output-parser - template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 10000 OUT.lines-retained --ignore-xml + template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 10000 $working-dir/corpus/parsed-clean.lines-retained.VERSION --ignore-xml error: there is a blank factor factorize in: clean-parsed-stem @@ -56,16 +55,8 @@ truecase template-if: input-truecaser IN.$input-extension OUT.$input-extension -model IN1.$input-extension template-if: output-truecaser IN.$output-extension OUT.$output-extension -model IN1.$output-extension parallelizable: yes -source-label - in: truecased-stem - out: source-labelled - default-name: corpus/labelled - pass-unless: source-labeller - template-if: source-labeller IN.$input-extension OUT.$input-extension - template-if: cat IN.$output-extension OUT.$output-extension - parallelizable: yes lowercase - in: source-labelled + in: truecased-stem out: lowercased-stem default-name: corpus/lowercased pass-unless: input-lowercaser output-lowercaser @@ -84,7 +75,7 @@ post-split-clean default-name: corpus/split-clean ignore-if: input-parser output-parser pass-unless: input-splitter output-splitter - template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 $max-sentence-length OUT.lines-retained + template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 $max-sentence-length $working-dir/corpus/split-clean.lines-retained.VERSION error: there is a blank factor post-split-clean-syntax in: split-stem @@ -92,7 +83,7 @@ post-split-clean-syntax default-name: corpus/split-clean ignore-unless: input-parser output-parser pass-unless: input-splitter output-splitter - template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 10000 OUT.lines-retained --ignore-xml + template: $moses-script-dir/training/clean-corpus-n.perl IN $input-extension $output-extension OUT 1 10000 $working-dir/corpus/parsed-clean.lines-retained.VERSION --ignore-xml error: there is a blank factor [RECASING] single @@ -105,9 +96,8 @@ tokenize train in: tokenized out: recase-config - template: $moses-script-dir/recaser/train-recaser.perl -train-script $TRAINING:script -dir OUT.model -corpus IN -scripts-root-dir $moses-script-dir -config OUT $recasing-settings + template: $moses-script-dir/recaser/train-recaser.perl -train-script $TRAINING:script -dir $working-dir/recasing/model.VERSION -corpus IN -scripts-root-dir $moses-script-dir -config OUT -ngram-count $lm-training default-name: recasing/moses.ini - tmp-name: recasing/model ignore-unless: EVALUATION:recaser error: cannot execute binary file @@ -116,14 +106,11 @@ consolidate in: CORPUS:clean-parsed-stem out: tokenized-stem default-name: truecaser/corpus - pass-unless: trainer template: $moses-script-dir/ems/support/consolidate-training-data.perl $input-extension $output-extension OUT IN - error: number of lines don't match train - in: tokenized-stem + in: tokenized-stem out: truecase-model rerun-on-change: trainer - pass-unless: trainer default-name: truecaser/truecase-model template: $trainer -model OUT.$input-extension -corpus IN.$input-extension ; $trainer -model OUT.$output-extension -corpus IN.$output-extension @@ -154,15 +141,10 @@ tokenize pass-unless: output-tokenizer template: $output-tokenizer < IN > OUT parallelizable: yes -mock-parse - in: tokenized-corpus - out: mock-parsed-corpus - default-name: lm/mock-parsed - pass-unless: mock-output-parser-lm - template: $mock-output-parser-lm < IN > OUT factorize - in: mock-parsed-corpus + in: tokenized-corpus out: factorized-corpus + rerun-on-change: TRAINING:output-factors default-name: lm/factored pass-unless: factors parallelizable: yes @@ -201,7 +183,6 @@ train rerun-on-change: lm-training order settings template: $lm-training -order $order $settings -text IN -lm OUT error: cannot execute binary file - error: unrecognised option randomize in: lm out: rlm @@ -228,7 +209,6 @@ binarize default-name: lm/binlm template: $lm-binarizer IN OUT error: set KENLM_MAX_ORDER to at least this value - final-model: yes [INTERPOLATED-LM] single tuning-from-sgm @@ -243,17 +223,12 @@ tokenize-tuning pass-unless: output-tokenizer template: $output-tokenizer < IN > OUT parallelizable: yes -mock-parse-tuning - in: tokenized-tuning - out: mock-parsed-tuning - default-name: lm/interpolate-tuning.mock-parsed - pass-unless: mock-output-parser-lm - template: $mock-output-parser-lm < IN > OUT factorize-tuning - in: mock-parsed-tuning + in: tokenized-tuning out: factorized-tuning + rerun-on-change: TRAINING:output-factors default-name: lm/interpolate-tuning.factored - pass-unless: TRAINING:output-factors + pass-unless: factors parallelizable: yes error: can't open error: incompatible number of words in factor @@ -280,7 +255,7 @@ split-tuning template: $output-splitter -model IN1.$output-extension < IN > OUT interpolate in: script split-tuning LM:lm - rerun-on-change: srilm-dir group weights + rerun-on-change: srilm-dir group out: lm default-name: lm/interpolated-lm randomize @@ -301,7 +276,6 @@ binarize rerun-on-change: lm default-name: lm/interpolated-binlm error: set kMaxOrder to at least this value - final-model: yes [MML] single tokenize-indomain-source in: raw-indomain-source @@ -412,7 +386,6 @@ build-domains default-name: model/domains ignore-unless: domain-features mml-filter-corpora template: $moses-script-dir/ems/support/build-domain-file-from-subcorpora.perl $input-extension IN > OUT - final-model: yes mml-score in: MML:model corpus domains out: mml-scores @@ -507,7 +480,6 @@ build-biconcor default-name: model/biconcor ignore-unless: biconcor error: usage - final-model: yes build-suffix-array in: corpus-mml-postfilter=OR=word-alignment corpus-mml-postfilter=OR=corpus-mml-prefilter=OR=corpus out: phrase-translation-table @@ -546,47 +518,25 @@ build-osm rerun-on-change: operation-sequence-model training-options script giza-settings operation-sequence-model-settings template: $moses-script-dir/OSM/OSM-Train.perl --corpus-f IN0.$input-extension --corpus-e IN0.$output-extension --alignment IN1.$alignment-symmetrization-method --order $operation-sequence-model-order --out-dir OUT --moses-src-dir $moses-src-dir --srilm-dir $srilm-dir $operation-sequence-model-settings default-name: model/OSM -build-transliteration-model - in: corpus word-alignment - out: transliteration-model - ignore-unless: transliteration-module - rerun-on-change: transliteration-module training-options script giza-settings - default-name: model/Transliteration - final-model: yes -build-translit-table - in: transliteration-model - out: transliteration-table - ignore-unless: in-decoding-transliteration - rerun-on-change: in-decoding-transliteration transliteration-module - default-name: model/transliteration-phrase-table - template: $moses-script-dir/Transliteration/in-decoding-transliteration.pl --moses-src-dir $moses-src-dir --external-bin-dir $external-bin-dir --transliteration-model-dir IN --input-extension $input-extension --output-extension $output-extension --transliteration-file $transliteration-file --out-file OUT extract-phrases in: corpus-mml-postfilter=OR=word-alignment scored-corpus out: extracted-phrases rerun-on-change: max-phrase-length translation-factors reordering-factors hierarchical-rule-set extract-settings training-options script use-ghkm domain-features baseline-extract lexicalized-reordering only-existence-matters: domain-features default-name: model/extract + ignore-if: suffix-array build-reordering in: extracted-phrases out: reordering-table ignore-unless: lexicalized-reordering rerun-on-change: lexicalized-reordering reordering-factors default-name: model/reordering-table - final-model: yes build-ttable in: extracted-phrases lexical-translation-table corpus-mml-prefilter=OR=corpus-mml-postfilter=OR=domains out: phrase-translation-table rerun-on-change: translation-factors hierarchical-rule-set score-settings training-options script EVALUATION:report-precision-by-coverage include-word-alignment-in-rules domain-features default-name: model/phrase-table - ignore-if: suffix-array mmsapt - final-model: yes -build-mmsapt - in: corpus-mml-postfilter=OR=word-alignment corpus-mml-postfilter=OR=corpus-mml-prefilter=OR=corpus - out: phrase-translation-table - ignore-unless: mmsapt - default-name: model/phrase-table-mmsapt - template: $moses-script-dir/training/build-mmsapt.perl --alignment IN.$alignment-symmetrization-method --corpus IN1 --f $input-extension --e $output-extension --dir OUT --settings '$mmsapt' - final-model: yes + ignore-if: suffix-array sigtest-filter-suffix-array in: corpus-mml-postfilter=OR=corpus-mml-prefilter=OR=corpus out: sigtest-filter-suffix-array @@ -602,22 +552,19 @@ sigtest-filter-suffix-array mv IN.${output-extension}.sa_offset OUT.${output-extension}.sa_offset ; \ mv IN.${output-extension}.sa_suffix OUT.${output-extension}.sa_suffix ignore-unless: sigtest-filter - final-model: yes sigtest-filter-ttable in: phrase-translation-table sigtest-filter-suffix-array out: sigtest-filter-phrase-translation-table default-name: model/phrase-table-sigtest-filter pass-unless: sigtest-filter - ignore-if: TRAINING:config - final-model: yes + ignore-if: TRAINING:config sigtest-filter-reordering in: reordering-table sigtest-filter-suffix-array out: sigtest-filter-reordering-table default-name: model/reordering-table-sigtest-filter pass-unless: sigtest-filter - ignore-if: TRAINING:config + ignore-if: TRAINING:config ignore-unless: lexicalized-reordering - final-model: yes build-generation in: corpus-mml-postfilter=OR=corpus-mml-prefilter=OR=corpus out: generation-table @@ -625,14 +572,12 @@ build-generation ignore-unless: generation-factors ignore-if: generation-corpus default-name: model/generation-table - final-model: yes build-generation-custom in: generation-corpus out: generation-table rerun-on-change: generation-factors generation-type training-options script generation-corpus ignore-unless: AND generation-factors generation-corpus default-name: model/generation-table - final-model: yes build-sparse in: corpus-mml-postfilter=OR=corpus-mml-prefilter=OR=corpus out: sparse @@ -641,13 +586,12 @@ build-sparse default-name: model/sparse-features template: $moses-script-dir/ems/support/build-sparse-features.perl IN $input-extension $output-extension OUT "$sparse-features" create-config - in: sigtest-filter-reordering-table sigtest-filter-phrase-translation-table transliteration-table generation-table sparse corpus-mml-prefilter=OR=corpus-mml-postfilter=OR=domains osm-model INTERPOLATED-LM:binlm LM:binlm + in: sigtest-filter-reordering-table sigtest-filter-phrase-translation-table generation-table sparse corpus-mml-prefilter=OR=corpus-mml-postfilter=OR=domains osm-model INTERPOLATED-LM:binlm LM:binlm out: config - ignore-if: use-hiero thot - rerun-on-change: decoding-steps alignment-factors translation-factors reordering-factors generation-factors lexicalized-reordering training-options script decoding-graph-backoff score-settings additional-ini mmsapt + ignore-if: use-hiero + rerun-on-change: decoding-steps alignment-factors translation-factors reordering-factors generation-factors lexicalized-reordering training-options script decoding-graph-backoff score-settings additional-ini default-name: model/moses.ini error: Unknown option - final-model: yes binarize-config in: config out: bin-config @@ -655,7 +599,6 @@ binarize-config rerun-on-change: config default-name: model/moses.bin.ini template: $binarize-all IN OUT -Binarizer $ttable-binarizer - final-model: yes hiero-compile-source-suffix-array in: corpus-mml-postfilter=OR=corpus-mml-prefilter=OR=corpus out: hiero-source-suffix-array @@ -700,18 +643,6 @@ hiero-create-config rerun-on-change: decoding-steps alignment-factors translation-factors reordering-factors generation-factors default-name: hiero-model/hiero.ini template: $hiero-util-dir/generate-ini.pl IN IN1 IN2 IN3 IN4 IN5 $hiero-max-phrase-length $hiero-max-nonterminals $hiero-max-phrase-span $hiero-min-gap-length $hiero-freq-rank1 $hiero-freq-rank2 < $GENERAL:hiero-template-ini > OUT -thot-build-ttable - in: corpus - out: thot-ttable - default-name: model/phrase-table-thot - rerun-on-change: input-extension output-extension - template: $thot/thot_tm_train -sdir $working-dir -s IN.$input-extension -t IN.$output-extension -o OUT -thot-create-config - in: thot-ttable LM:lm - out: config - ignore-unless: thot - default-name: model/thot.ini - template: $thot/thot_gen_cfg_file IN1/lm_desc IN/tm_desc > OUT [TUNING] single input-from-sgm @@ -738,32 +669,17 @@ tokenize-input-devtest pass-unless: input-tokenizer ignore-unless: use-mira template: $input-tokenizer < IN > OUT -mock-parse-input - in: tokenized-input - out: mock-parsed-input - default-name: tuning/input.mock-parsed - pass-unless: mock-input-parser-devtesteval - template: $mock-input-parser-devtesteval < IN > OUT -mock-parse-input-devtest - in: tokenized-input-devtest - out: mock-parsed-input-devtest - default-name: tuning/input.devtest.mock-parsed - pass-unless: mock-input-parser-devtesteval - ignore-unless: use-mira - template: $mock-input-parser-devtesteval < IN > OUT parse-input - in: mock-parsed-input + in: tokenized-input out: parsed-input default-name: tuning/input.parsed pass-unless: input-parser - pass-if: skip-parse-input-devtesteval mock-input-parser-devtesteval template: $input-parser < IN > OUT parse-input-devtest - in: mock-parsed-input-devtesteval + in: tokenized-input-devtest out: parsed-input-devtest default-name: tuning/input.devtest.parsed pass-unless: input-parser - pass-if: skip-parse-input-devtesteval mock-input-parser-devtesteval ignore-unless: use-mira template: $input-parser < IN > OUT parse-relax-input @@ -771,16 +687,14 @@ parse-relax-input out: parse-relaxed-input default-name: tuning/input.parse-relaxed pass-unless: input-parse-relaxer - pass-if: skip-parse-input-devtesteval mock-input-parser-devtesteval - template: $input-parse-relaxer < IN > OUT + template: $input-parse-relaxer < IN.$input-extension > OUT.$input-extension parse-relax-input-devtest in: parsed-input-devtest out: parse-relaxed-input-devtest default-name: tuning/input.devtest.parse-relaxed pass-unless: input-parse-relaxer - pass-if: skip-parse-input-devtesteval mock-input-parser-devtesteval ignore-unless: use-mira - template: $input-parse-relaxer < IN > OUT + template: $input-parse-relaxer < IN.$input-extension > OUT.$input-extension factorize-input in: parse-relaxed-input out: factorized-input @@ -798,29 +712,15 @@ factorize-input-devtest ignore-unless: use-mira error: can't open error: incompatible number of words in factor -source-label-input - in: factorized-input - out: source-labelled-input - default-name: tuning/input.labelled - pass-unless: source-labeller - template-if: source-labeller IN OUT - parallelizable: yes -source-label-input-devtest - in: factorized-input-devtest - out: source-labelled-input-devtest - default-name: tuning/input.devtest.labelled - pass-unless: source-labeller - template-if: source-labeller IN OUT - parallelizable: yes lowercase-input - in: source-labelled-input + in: factorized-input out: truecased-input default-name: tuning/input.lc pass-unless: input-lowercaser ignore-if: input-truecaser template: $input-lowercaser < IN > OUT lowercase-input-devtest - in: source-labelled-input-devtest + in: factorized-input-devtest out: truecased-input-devtest default-name: tuning/input.devtest.lc pass-unless: input-lowercaser @@ -828,14 +728,14 @@ lowercase-input-devtest ignore-if: input-truecaser template: $input-lowercaser < IN > OUT truecase-input - in: source-labelled-input TRUECASER:truecase-model + in: factorized-input TRUECASER:truecase-model out: truecased-input rerun-on-change: input-truecaser default-name: tuning/input.tc ignore-unless: input-truecaser template: $input-truecaser -model IN1.$input-extension < IN > OUT truecase-input-devtest - in: source-labelled-input-devtest TRUECASER:truecase-model + in: factorized-input-devtest TRUECASER:truecase-model out: truecased-input-devtest rerun-on-change: input-truecaser default-name: tuning/input.devtest.tc @@ -882,20 +782,8 @@ tokenize-reference-devtest ignore-unless: use-mira multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-tokenizer < IN > OUT -mock-parse-reference - in: tokenized-reference - out: mock-parsed-reference - default-name: tuning/reference.mock-parsed - pass-unless: mock-output-parser-references - template: $mock-output-parser-references < IN > OUT -mock-parse-reference-devtest - in: tokenized-input-devtest - out: mock-parsed-reference-devtest - default-name: tuning/reference.devtest.mock-parsed - pass-unless: mock-output-parser-references - template: $mock-output-parser-references < IN > OUT lowercase-reference - in: mock-parsed-reference + in: tokenized-reference out: truecased-reference default-name: tuning/reference.lc pass-unless: output-lowercaser @@ -903,7 +791,7 @@ lowercase-reference multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-lowercaser < IN > OUT lowercase-reference-devtest - in: mock-parsed-reference-devtest + in: tokenized-reference-devtest out: truecased-reference-devtest default-name: tuning/reference.devtest.lc pass-unless: output-lowercaser @@ -912,7 +800,7 @@ lowercase-reference-devtest multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-lowercaser < IN > OUT truecase-reference - in: mock-parsed-reference TRUECASER:truecase-model + in: tokenized-reference TRUECASER:truecase-model out: truecased-reference rerun-on-change: output-truecaser default-name: tuning/reference.tc @@ -920,7 +808,7 @@ truecase-reference multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-truecaser -model IN1.$output-extension < IN > OUT truecase-reference-devtest - in: mock-parsed-reference-devtest TRUECASER:truecase-model + in: tokenized-reference-devtest TRUECASER:truecase-model out: truecased-reference-devtest rerun-on-change: output-truecaser default-name: tuning/reference.devtest.tc @@ -943,12 +831,13 @@ split-reference-devtest multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-splitter -model IN1.$output-extension < IN > OUT filter - in: input TRAINING:sigtest-filter-phrase-translation-table TRAINING:sigtest-filter-reordering-table TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus-mml-postfilter=OR=TRAINING:domains TRAINING:transliteration-table + in: input TRAINING:sigtest-filter-phrase-translation-table TRAINING:sigtest-filter-reordering-table TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus-mml-postfilter=OR=TRAINING:domains out: filtered-dir default-name: tuning/filtered rerun-on-change: filter-settings ttable-binarizer ignore-if: TRAINING:binarize-all error: already exists. Please delete + error: No phrases found in filter-devtest in: input-devtest TRAINING:sigtest-filter-phrase-translation-table TRAINING:sigtest-filter-reordering-table out: filtered-dir-devtest @@ -976,21 +865,12 @@ tune ignore-if: use-hiero qsub-script: yes default-name: tuning/moses.ini - tmp-name: tuning/tmp - final-model: yes rerun-on-change: decoder-settings tuning-settings nbest lambda async not-error: trans: No such file or directory -thot-tune - in: TRAINING:config input reference - out: config-with-reused-weights - ignore-unless: thot - tmp-name: tuning/thot.tmp - default-name: tuning/thot.tuned.ini - template: mkdir -p TMP/home ; mkdir -p TMP/tdir ; mkdir -p TMP/sdir ; HOME=TMP/home $thot/thot_smt_tune -tdir TMP/tdir -sdir TMP/sdir -c IN -s IN1 -t IN2 -o OUT apply-weights in: TRAINING:bin-config weight-config out: config-with-reused-weights - ignore-if: use-hiero thot + ignore-if: use-hiero default-name: tuning/moses.tuned.ini template: $moses-script-dir/ems/support/substitute-weights.perl IN IN1 OUT error: cannot open @@ -1028,26 +908,18 @@ tokenize-input default-name: evaluation/input.tok pass-unless: input-tokenizer template: $input-tokenizer < IN > OUT -mock-parse-input - in: tokenized-input - out: mock-parsed-input - default-name: evaluation/input.mock-parsed - pass-unless: mock-input-parser-devtesteval - template: $mock-input-parser-devtesteval < IN > OUT parse-input - in: mock-parsed-input + in: tokenized-input out: parsed-input default-name: evaluation/input.parsed pass-unless: input-parser - pass-if: skip-parse-input-devtesteval mock-input-parser-devtesteval template: $input-parser < IN > OUT parse-relax-input in: parsed-input out: parse-relaxed-input - default-name: evaluation/input.parse-relaxed + default-name: tuning/input.parse-relaxed pass-unless: input-parse-relaxer - pass-if: skip-parse-input-devtesteval mock-input-parser-devtesteval - template: $input-parse-relaxer < IN > OUT + template: $input-parse-relaxer < IN.$input-extension > OUT.$input-extension factorize-input in: parse-relaxed-input out: factorized-input @@ -1056,24 +928,15 @@ factorize-input pass-unless: TRAINING:input-factors error: can't open error: incompatible number of words in factor - -source-label-input - in: factorized-input - out: source-labelled-input - default-name: evaluation/input.labelled - pass-unless: source-labeller - template-if: source-labeller IN OUT - parallelizable: yes - lowercase-input - in: source-labelled-input + in: factorized-input out: truecased-input default-name: evaluation/input.lc pass-unless: input-lowercaser ignore-if: input-truecaser template: $input-lowercaser < IN > OUT truecase-input - in: source-labelled-input TRUECASER:truecase-model + in: factorized-input TRUECASER:truecase-model out: truecased-input default-name: evaluation/input.tc rerun-on-change: input-truecaser @@ -1086,8 +949,8 @@ split-input pass-unless: input-splitter template: $input-splitter -model IN1.$input-extension < IN > OUT filter - in: input TRAINING:sigtest-filter-phrase-translation-table TRAINING:sigtest-filter-reordering-table TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus-mml-postfilter=OR=TRAINING:domains TRAINING:transliteration-table - out: filtered-dir + in: input TRAINING:sigtest-filter-phrase-translation-table TRAINING:sigtest-filter-reordering-table TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus-mml-postfilter=OR=TRAINING:domains + out: filtered-dir default-name: evaluation/filtered rerun-on-change: filter-settings report-precision-by-coverage ttable-binarizer pass-if: TRAINING:binarize-all @@ -1097,18 +960,17 @@ apply-filter in: filtered-dir TRAINING:config TUNING:config-with-reused-weights out: filtered-config default-name: evaluation/filtered.ini - ignore-if: TRAINING:binarize-all thot + ignore-if: TRAINING:binarize-all template: $moses-script-dir/ems/support/substitute-filtered-tables-and-weights.perl IN/moses.ini IN1 IN2 OUT decode in: TUNING:config-with-reused-weights input filtered-config out: system-output default-name: evaluation/output qsub-script: yes - ignore-if: use-hiero thot - rerun-on-change: decoder decoder-settings nbest report-segmentation report-precision-by-coverage analyze-search-graph wade TRAINING:post-decoding-transliteration + ignore-if: use-hiero + rerun-on-change: decoder decoder-settings nbest report-segmentation report-precision-by-coverage analyze-search-graph wade error: Translation was not performed correctly not-error: trans: No such file or directory - final-model: yes hiero-decode in: TUNING:hiero-config-with-reused-weights input out: system-output @@ -1117,20 +979,6 @@ hiero-decode ignore-unless: use-hiero template: $hiero-parallelizer -e OUT.edir -r -- $hiero-decoder -c IN < IN1 > OUT rerun-on-change: hiero-decoder -thot-filter - in: TUNING:config-with-reused-weights input - out: filtered-config - ignore-unless: thot - default-name: evaluation/filtered - tmp-name: evaluation/filtered-tmp - template: mkdir -p TMP/home ; mkdir -p TMP/tdir ; mkdir -p TMP/sdir ; HOME=TMP/home $thot/thot_prepare_sys_for_test -sdir TMP/sdir -tdir TMP/tdir -t IN1 -c IN/tuned_for_dev.cfg -o OUT ; cp OUT/lm/main/* OUT/lm -thot-decode - in: input filtered-config - out: system-output - ignore-unless: thot - default-name: evaluation/output - template: $thot/thot_decoder -sdir $working-dir -c IN1/test_specific.cfg -t IN > OUT - not-error: Error in word penalty model file remove-markup in: system-output out: cleaned-output @@ -1138,21 +986,15 @@ remove-markup pass-if: TRAINING:hierarchical-rule-set pass-unless: report-segmentation template: $moses-script-dir/ems/support/remove-segmentation-markup.perl < IN > OUT -post-decoding-transliteration - in: cleaned-output system-output TRAINING:transliteration-model INTERPOLATED-LM:binlm=OR=LM:binlm - out: transliterated-output - default-name: evaluation/transliterated - pass-unless: TRAINING:post-decoding-transliteration - template: $moses-script-dir/Transliteration/post-decoding-transliteration.pl --moses-src-dir $moses-src-dir --external-bin-dir $external-bin-dir --transliteration-model-dir IN2 --input-extension $input-extension --output-extension $output-extension --language-model IN3 --input-file IN0 --output-file OUT --oov-file IN1.oov --decoder $decoder recase-output - in: transliterated-output RECASING:recase-config + in: cleaned-output RECASING:recase-config out: recased-output default-name: evaluation/recased pass-unless: recaser ignore-if: output-truecaser template: $recaser -moses $RECASING:decoder -in IN -model IN1 > OUT detruecase-output - in: transliterated-output + in: cleaned-output out: recased-output default-name: evaluation/truecased ignore-unless: output-truecaser @@ -1163,7 +1005,6 @@ detokenize-output default-name: evaluation/detokenized pass-unless: detokenizer template: $detokenizer < IN > OUT - final-model: yes wrap in: detokenized-output out: wrapped-output @@ -1171,7 +1012,6 @@ wrap rerun-on-change: wrapping-frame use-hiero template: $wrapping-script $wrapping-frame < IN > OUT error: Use of uninitialized value in pattern match - final-model: yes reference-from-sgm in: reference-sgm input-sgm out: raw-reference @@ -1184,14 +1024,8 @@ tokenize-reference pass-unless: output-tokenizer multiref: $moses-script-dir/ems/support/run-command-on-multiple-refsets.perl template: $output-tokenizer < IN > OUT -mock-parse-reference - in: tokenized-reference - out: mock-parsed-reference - default-name: evaluation/reference.mock-parsed - pass-unless: mock-output-parser-references - template: $mock-output-parser-references < IN > OUT lowercase-reference - in: mock-parsed-reference + in: tokenized-reference out: reference default-name: evaluation/reference pass-unless: output-lowercaser @@ -1213,7 +1047,6 @@ nist-bleu rerun-on-change: nist-bleu error: Illegal division by zero template: $nist-bleu -s $input-sgm -r IN1 -t IN > OUT - final-model: yes nist-bleu-c in: wrapped-output reference-sgm out: nist-bleu-c-score @@ -1222,7 +1055,6 @@ nist-bleu-c rerun-on-change: nist-bleu-c error: Illegal division by zero template: $nist-bleu-c -c -s $input-sgm -r IN1 -t IN > OUT - final-model: yes ibm-bleu in: wrapped-output reference-sgm out: ibm-bleu-score @@ -1230,7 +1062,6 @@ ibm-bleu ignore-unless: ibm-bleu rerun-on-change: ibm-bleu template: $ibm-bleu -ci -s $input-sgm -r IN1 -t IN > OUT - final-model: yes ibm-bleu-c in: wrapped-output reference-sgm out: ibm-bleu-c-score @@ -1238,7 +1069,6 @@ ibm-bleu-c ignore-unless: ibm-bleu-c rerun-on-change: ibm-bleu-c template: $ibm-bleu-c -s $input-sgm -r IN1 -t IN > OUT - final-model: yes bolt-bleu in: detokenized-output out: bolt-bleu-score @@ -1246,7 +1076,6 @@ bolt-bleu ignore-unless: bolt-bleu rerun-on-change: bolt-bleu template: $bolt-bleu IN > OUT - final-model: yes bolt-bleu-c in: detokenized-output out: bolt-bleu-c-score @@ -1254,15 +1083,13 @@ bolt-bleu-c ignore-unless: bolt-bleu-c rerun-on-change: bolt-bleu-c template: $bolt-bleu-c IN > OUT - final-model: yes multi-bleu - in: transliterated-output tokenized-reference + in: cleaned-output tokenized-reference out: multi-bleu-score default-name: evaluation/multi-bleu ignore-unless: multi-bleu rerun-on-change: multi-bleu template: $multi-bleu IN1 < IN > OUT - final-model: yes multi-bleu-c in: recased-output tokenized-reference out: multi-bleu-c-score @@ -1270,14 +1097,12 @@ multi-bleu-c ignore-unless: multi-bleu-c rerun-on-change: multi-bleu-c template: $multi-bleu-c IN1 < IN > OUT - final-model: yes ter in: wrapped-output reference-sgm out: ter-score default-name: evaluation/detokenized.sgm.TER ignore-unless: ter rerun-on-change: ter - final-model: yes wer in: recased-output reference out: wer-score @@ -1285,36 +1110,32 @@ wer ignore-unless: wer rerun-on-change: wer template: $wer IN IN1 > OUT - final-model: yes meteor - in: transliterated-output reference + in: cleaned-output reference out: meteor-score default-name: evaluation/meteor ignore-unless: meteor rerun-on-change: meteor template: $meteor IN IN1 $meteor-params > OUT - final-model: yes analysis in: recased-output reference input out: analysis default-name: evaluation/analysis ignore-if: report-precision-by-coverage ignore-unless: analysis - rerun-on-change: analyze-search-graph + rerun-on-change: analyze-search-graph analysis-coverage in: input TRAINING:corpus-mml-postfilter=OR=TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus TRAINING:sigtest-filter-phrase-translation-table out: analysis-coverage default-name: evaluation/analysis ignore-unless: AND analysis analyze-coverage rerun-on-change: score-settings - final-model: yes analysis-precision in: recased-output reference input TRAINING:corpus-mml-postfilter=OR=TRAINING:corpus-mml-prefilter=OR=TRAINING:corpus TRAINING:sigtest-filter-phrase-translation-table analysis-coverage out: analysis default-name: evaluation/analysis ignore-unless: AND analysis analyze-coverage report-precision-by-coverage - rerun-on-change: precision-by-coverage-base - final-model: yes + rerun-on-change: precision-by-coverage-base [REPORTING] single report |