Merge branch 'master' of https://github.com/marian-nmt/marian-examples

1 files changed, 142 insertions, 56 deletions

diff --git a/training-basics-spm/README.md b/training-basics-spm/README.md
index d4bb424..2563775 100644
--- a/training-basics-spm/README.md
+++ b/training-basics-spm/README.md
@@ -1,8 +1,12 @@
-# Example for training with Marian and SentencePiece
+# Example for Training with Marian and SentencePiece
 
-## Building Marian with SentencePiece support
+In this example, we modify the Romanian-English example from `examples/training-basics` to use Taku Kudo's 
+[SentencePiece](https://github.com/google/sentencepiece) instead of a complicated pre/prost-processing pipeline. 
+We also replace the evaluation scripts with Matt Post's [SacreBLEU](https://github.com/mjpost/sacreBLEU). Both tools greatly simplify the training and evaluation process by providing ways to have reversible hidden preprocessing and repeatable evaluation. 
 
-Since version 1.7.0, Marian has support for (SentencePiece)[https://github.com/google/sentencepiece],
+## Building Marian with SentencePiece Support
+
+Since version 1.7.0, Marian has built-in support for SentencePiece,
 but this needs to be enabled at compile-time. We decided to make the compilation of SentencePiece
 optional as SentencePiece has a number of dependencies - especially Google's Protobuf - that
 are potentially non-trivial to install.
@@ -37,6 +41,7 @@ With these dependencies met, you can compile Marian as follows:
 git clone https://github.com/marian-nmt/marian
 cd marian
 mkdir build
+cd build
 cmake .. -DCMAKE_BUILD_TYPE=Release -DUSE_SENTENCEPIECE=ON
 make -j 8
 ```
@@ -47,7 +52,7 @@ To test if `marian` has been compiled with SentencePiece support run
 ./marian --help |& grep sentencepiece
 ```
 
-which should display the following new options
+which should display the following new options:
 
 ```
   --sentencepiece-alphas VECTOR ...     Sampling factors for SentencePieceVocab; i-th factor corresponds to i-th vocabulary
@@ -55,7 +60,7 @@ which should display the following new options
   --sentencepiece-max-lines UINT=10000000
 ```
 
-##
+## Execute the Example
 
 Files and scripts in this folder have been adapted from the Romanian-English
 sample from https://github.com/rsennrich/wmt16-scripts. We also add the
@@ -65,83 +70,164 @@ http://www.aclweb.org/anthology/W16-2323. The resulting system should be
 competitive or even slightly better than reported in the Edinburgh WMT2016
 paper.
 
-To execute the complete example type:
+Assuming you one GPU, to execute the complete example type:
 
 ```
 ./run-me.sh
 ```
 
-which downloads the Romanian-English training files and preprocesses them (tokenization,
-truecasing, segmentation into subwords units).
+which downloads the Romanian-English training files and concatenates them into training files. 
+No preprocessing is required as the Marian command will train a SentencePiece vocabulary from
+the raw text. Next the translation model will be trained and after convergence, the dev and test
+sets are translated and evaluated with sacreBLEU.
 
-To use with a different GPU than device 0 or more GPUs (here 0 1 2 3) type the command below.
-Training time on 1 NVIDIA GTX 1080 GPU should be roughly 24 hours.
+To use with a different GPUs than device 0 or more GPUs (here 0 1 2 3) use the command below:
 
 ```
 ./run-me.sh 0 1 2 3
 ```
 
-Next it executes a training run with `marian`:
+## Step-by-step Walkthrough
+
+In this section we repeat the content from the above `run-me.sh` script with explanations. You should be able to copy and paste the commands and follow through all the steps. 
+
+We assume you are running these commands from the examples directory of the main Marian directory tree `marian/examples/training-basics-spm` and that the Marian binaries have been compiled in `marian/build`. The localization of the Marian binary relative to the current directory is therefore `../../build/marian`.
+
+### Preparing the test and validation sets
+
+```
+# get our fork of sacrebleu
+git clone https://github.com/marian-nmt/sacreBLEU.git sacreBLEU
+
+# create dev set
+sacreBLEU/sacrebleu.py -t wmt16/dev -l ro-en --echo src > data/newsdev2016.ro
+sacreBLEU/sacrebleu.py -t wmt16/dev -l ro-en --echo ref > data/newsdev2016.en
+
+# create test set
+sacreBLEU/sacrebleu.py -t wmt16 -l ro-en --echo src > data/newstest2016.ro
+sacreBLEU/sacrebleu.py -t wmt16 -l ro-en --echo ref > data/newstest2016.en
+```
+
+### Downloading the training files
+
+```
+# change into data directory
+cd data
+
+# get En-Ro training data for WMT16
+wget -nc http://www.statmt.org/europarl/v7/ro-en.tgz
+wget -nc http://opus.lingfil.uu.se/download.php?f=SETIMES2/en-ro.txt.zip -O SETIMES2.ro-en.txt.zip
+wget -nc http://data.statmt.org/rsennrich/wmt16_backtranslations/ro-en/corpus.bt.ro-en.en.gz
+wget -nc http://data.statmt.org/rsennrich/wmt16_backtranslations/ro-en/corpus.bt.ro-en.ro.gz
+
+# extract data
+tar -xf ro-en.tgz
+unzip SETIMES2.ro-en.txt.zip
+gzip -d corpus.bt.ro-en.en.gz corpus.bt.ro-en.ro.gz
+
+# create corpus files
+cat europarl-v7.ro-en.en SETIMES2.en-ro.en corpus.bt.ro-en.en > corpus.en
+cat europarl-v7.ro-en.ro SETIMES2.en-ro.ro corpus.bt.ro-en.ro > corpus.ro
+
+# clean
+rm ro-en.tgz SETIMES2.* corpus.bt.* europarl-*
+
+# change back into main directory
+cd ..
+```
+
+### Normalization of Romanian diacritics with SentencePiece
+
+It seems that the training data is quite noisy and multiple similar characters are used in place of the one correct character.
+Barry Haddow from Edinburgh who created the original Python scripts noticed that removing diacritics on the Romanian side leads to a significant improvment in translation quality. And indeed we saw gains of up to 2 BLEU points due to normalization versus unnormalized text. 
+
+SentencePiece allows to specify normalization or replacement tables for character sequences. These replacements are applied before tokenization/segmentation and included in the SentencePiece model. Based on these preprocessing scripts from `test`, we manually create a tab-separated normalization rule file `data/norm_romanian.tsv` looking like this (see the [SentencePiece documentation on normalization](https://) for details):
+
+```
+015E    53 # Ş => S
+015F    73 # ş => s
+0162    54 # Ţ => T
+0163    74 # ţ => t
+0218    53 # Ș => S
+0219    73 # ș => s
+021A    54 # Ț => T
+021B    74 # ț => t
+0102    41 # Ă => A
+0103    61 # ă => a
+00C2    41 # Â => A
+00E2    61 # â => a
+00CE    49 # Î => I
+00EE    69 # î => i
+```
+
+### Training the NMT model
+
+Next, we execute a training run with `marian`. Note how the training command is called passing the 
+raw training and validation data into Marian. A single joint SentencePiece model will be saved to 
+`model/vocab.roen.spm`. The `*.spm` suffix is required and tells Marian to train a SentencePiece 
+vocabulary. When the same vocabulary file is specified multiple times - like in this example - a single
+vocabulary is built for the union of the corresponding training files. This also enables us to use
+tied embeddings (`--tied-embeddings-all`).
+
+We can pass the Romanian-specific normalizaton rules via the `--sentencepiece-options` command line
+argument. The values of this option are passed on to the SentencePiece trainer, note the required single
+quotes around the SentencePiece options: `--sentencepiece-options '--normalization_rule_tsv=data/norm_romanian.tsv'`.
+
+Another new feature is the `bleu-detok` validation metric. When used with SentencePiece this should
+give you in-training BLEU scores that are very close to sacreBLEU's scores. Differences may appear 
+if unexpected SentencePiece normalization rules are used. You should still report only official
+sacreBLEU scores for publications.
+
+We are training of four GPUs defined with `--devices 0 1 2 3`. Change this to the required number of GPUs.
 
 ```
 ../../build/marian \
-    --devices $GPUS \
-    --type amun \
+    --devices 0 1 2 3 \
+    --type s2s \
     --model model/model.npz \
-    --train-sets data/corpus.bpe.ro data/corpus.bpe.en \
-    --vocabs model/vocab.ro.yml model/vocab.en.yml \
-    --dim-vocabs 66000 50000 \
-    --mini-batch-fit -w 3000 \
-    --layer-normalization --dropout-rnn 0.2 --dropout-src 0.1 --dropout-trg 0.1 \
-    --early-stopping 5 \
+    --train-sets data/corpus.ro data/corpus.en \
+    --vocabs model/vocab.roen.spm model/vocab.roen.spm \
+    --sentencepiece-options '--normalization_rule_tsv=data/norm_romanian.tsv' \
+    --dim-vocabs 32000 32000 \
+    --mini-batch-fit -w 5000 \
+    --layer-normalization --tied-embeddings-all \
+    --dropout-rnn 0.2 --dropout-src 0.1 --dropout-trg 0.1 \
+    --early-stopping 5 --max-length 100 \
     --valid-freq 10000 --save-freq 10000 --disp-freq 1000 \
-    --valid-metrics cross-entropy translation \
-    --valid-sets data/newsdev2016.bpe.ro data/newsdev2016.bpe.en \
-    --valid-script-path ./scripts/validate.sh \
-    --log model/train.log --valid-log model/valid.log \
+    --cost-type ce-mean-words --valid-metrics ce-mean-words bleu-detok \
+    --valid-sets data/newsdev2016.ro data/newsdev2016.en \
+    --log model/train.log --valid-log model/valid.log --tempdir model \
     --overwrite --keep-best \
     --seed 1111 --exponential-smoothing \
-    --normalize=1 --beam-size 12 --quiet-translation
+    --normalize=0.6 --beam-size=6 --quiet-translation
 ```
 
-After training (the training should stop if cross-entropy on the validation set
-stops improving) the model with the highest translation validation score is used
+The training should stop if cross-entropy on the validation set
+stops improving.
+
+### Translating the test and validation sets with evaluation
+
+After training, the model with the highest translation validation score is used
 to translate the WMT2016 dev set and test set with `marian-decoder`:
 
 ```
-cat data/newsdev2016.bpe.ro \
-    | ../../build/marian-decoder -c  model/model.npz.best-translation.npz.decoder.yml -d $GPUS \
-      -b 12 -n1 --mini-batch 64 --maxi-batch 10 --maxi-batch-sort src -w 2500 \
-    | sed 's/\@\@ //g' \
-    | ../tools/moses-scripts/scripts/recaser/detruecase.perl \
-    | ../tools/moses-scripts/scripts/tokenizer/detokenizer.perl -l en \
-    > data/newsdev2016.ro.output
-```
-after which BLEU scores for the dev and test set are reported. Results should
-be somewhere in the area of:
+# translate dev set
+cat data/newsdev2016.ro \
+    | ../../build/marian-decoder -c model/model.npz.best-bleu-detok.npz.decoder.yml -d 0 1 2 3 -b 6 -n0.6 \
+      --mini-batch 64 --maxi-batch 100 --maxi-batch-sort src > data/newsdev2016.ro.output
 
+# translate test set
+cat data/newstest2016.ro \
+    | ../../build/marian-decoder -c model/model.npz.best-bleu-detok.npz.decoder.yml -d 0 1 2 3 -b 6 -n0.6 \
+      --mini-batch 64 --maxi-batch 100 --maxi-batch-sort src > data/newstest2016.ro.output
 ```
-newsdev2016:
-BLEU = 35.88, 67.4/42.3/28.8/20.2 (BP=1.000, ratio=1.012, hyp_len=51085, ref_len=50483)
-
-newstest2016:
-BLEU = 34.53, 66.0/40.7/27.5/19.2 (BP=1.000, ratio=1.015, hyp_len=49258, ref_len=48531)
+after which BLEU scores for the dev and test set are reported.
 ```
-
-## Custom validation script
-
-The validation script `scripts/validate.sh` is a quick example how to write a
-custom validation script. The training pauses until the validation script
-finishes executing. A validation script should not output anything to `stdout`
-apart from the final single score (last line):
-
+# calculate bleu scores on dev and test set
+sacreBLEU/sacrebleu.py -t wmt16/dev -l ro-en < data/newsdev2016.ro.output
+sacreBLEU/sacrebleu.py -t wmt16 -l ro-en < data/newstest2016.ro.output
+```
+You should see results somewhere in the area of:
 ```
-#!/bin/bash
 
-cat $1 \
-    | sed 's/\@\@ //g' \
-    | ../tools/moses-scripts/scripts/recaser/detruecase.perl \
-    | ../tools/moses-scripts/scripts/tokenizer/detokenize.perl -l en \
-    | ../tools/moses-scripts/scripts/generic/multi-bleu-detok.perl data/newsdev2016.en \
-    | sed -r 's/BLEU = ([0-9.]+),.*/\1/'
 ```