Fix multilingual evaluation bug (#592)

Summary:
Pull Request resolved: https://github.com/pytorch/translate/pull/592

Fix bug reported at
https://github.com/pytorch/fairseq/commit/9c3bb5c6d6c7d6442a28ccb8a81b2fc4e5782ace#r34181600

D15682169 breaks the multilingual translation generation.

Reviewed By: dpacgopinath

Differential Revision: D16147454

fbshipit-source-id: e0cf4d32f362190a0542fa0160f65a2a207ca3fa

fairseq/tasks/multilingual_translation.py

@@ -100,19 +100,23 @@ class MultilingualTranslationTask(FairseqTask):
     def __init__(self, args, dicts, training):
         super().__init__(args)
         self.dicts = dicts
+        self.training = training
+        if training:
             self.lang_pairs = args.lang_pairs
+            args.source_lang, args.target_lang = args.lang_pairs[0].split('-')
+        else:
+            self.lang_pairs = ['{}-{}'.format(args.source_lang, args.target_lang)]
         # eval_lang_pairs for multilingual translation is usually all of the
         # lang_pairs. However for other multitask settings or when we want to
         # optimize for certain languages we want to use a different subset. Thus
         # the eval_lang_pairs class variable is provided for classes that extend
         # this class.
-        self.eval_lang_pairs = args.lang_pairs
+        self.eval_lang_pairs = self.lang_pairs
         # model_lang_pairs will be used to build encoder-decoder model pairs in
         # models.build_model(). This allows multitask type of sub-class can
         # build models other than the input lang_pairs
-        self.model_lang_pairs = copy.copy(args.lang_pairs)
+        self.model_lang_pairs = self.lang_pairs
         self.langs = list(dicts.keys())
-        self.training = training
 
     @classmethod
     def setup_task(cls, args, **kwargs):
@@ -136,10 +140,8 @@ class MultilingualTranslationTask(FairseqTask):
         sorted_langs = sorted(list({x for lang_pair in args.lang_pairs for x in lang_pair.split('-')}))
         if args.source_lang is not None or args.target_lang is not None:
             training = False
-            args.lang_pairs = ['{}-{}'.format(args.source_lang, args.target_lang)]
         else:
             training = True
-            args.source_lang, args.target_lang = args.lang_pairs[0].split('-')
 
         # load dictionaries
         dicts = OrderedDict()
@@ -224,7 +226,7 @@ class MultilingualTranslationTask(FairseqTask):
         self.datasets[split] = RoundRobinZipDatasets(
             OrderedDict([
                 (lang_pair, language_pair_dataset(lang_pair))
-                for lang_pair in self.args.lang_pairs
+                for lang_pair in self.lang_pairs
             ]),
             eval_key=None if self.training else "%s-%s" % (self.args.source_lang, self.args.target_lang),
         )
@@ -272,7 +274,7 @@ class MultilingualTranslationTask(FairseqTask):
     def train_step(self, sample, model, criterion, optimizer, ignore_grad=False):
         model.train()
         agg_loss, agg_sample_size, agg_logging_output = 0., 0., {}
-        for lang_pair in self.args.lang_pairs:
+        for lang_pair in self.model_lang_pairs:
             if sample[lang_pair] is None or len(sample[lang_pair]) == 0:
                 continue
             loss, sample_size, logging_output = criterion(model.models[lang_pair], sample[lang_pair])

fairseq/tasks/semisupervised_translation.py

@@ -124,7 +124,7 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
                 "%s-%s" % (tgt, tgt)
                 for tgt in {lang_pair.split('-')[1] for lang_pair in args.lang_pairs}
             ]
-            self.model_lang_pairs += denoising_lang_pairs
+            self.model_lang_pairs = self.model_lang_pairs + denoising_lang_pairs
         self.backtranslate_datasets = {}
         self.backtranslators = {}
 
@@ -164,7 +164,7 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
         # load parallel datasets
         src_datasets, tgt_datasets = {}, {}
         if (self.lambda_parallel > 0.0 or self.lambda_parallel_steps is not None or not split.startswith("train")):
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 src, tgt = lang_pair.split('-')
                 if split_exists(split, src, tgt, src):
                     prefix = os.path.join(data_path, '{}.{}-{}.'.format(split, src, tgt))
@@ -181,7 +181,7 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
         # back translation datasets
         backtranslate_datasets = {}
         if (self.lambda_otf_bt > 0.0 or self.lambda_otf_bt_steps is not None) and split.startswith("train"):
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 src, tgt = lang_pair.split('-')
                 if not split_exists(split, tgt, None, tgt):
                     raise FileNotFoundError('Dataset not found: backtranslation {} ({})'.format(split, data_path))
@@ -229,7 +229,7 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
         # denoising autoencoder
         noising_datasets = {}
         if (self.lambda_denoising > 0.0 or self.lambda_denoising_steps is not None) and split.startswith("train"):
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 _, tgt = lang_pair.split('-')
                 if not split_exists(split, tgt, None, tgt):
                     continue
@@ -305,7 +305,7 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
         # create SequenceGenerator for each model that has backtranslation dependency on it
         self.sequence_generators = {}
         if (self.lambda_otf_bt > 0.0 or self.lambda_otf_bt_steps is not None) and self.training:
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 src, tgt = lang_pair.split('-')
                 key = '{}-{}'.format(tgt, src)
                 self.sequence_generators[key] = SequenceGenerator(
@@ -350,16 +350,16 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
             agg_logging_output[logging_output_key] = logging_output
 
         if self.lambda_parallel > 0.0:
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 forward_backward(model.models[lang_pair], sample[lang_pair], lang_pair, self.lambda_parallel)
 
         if self.lambda_otf_bt > 0.0:
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 sample_key = _get_bt_dataset_key(lang_pair)
                 forward_backward(model.models[lang_pair], sample[sample_key], sample_key, self.lambda_otf_bt)
 
         if self.lambda_denoising > 0.0:
-            for lang_pair in self.args.lang_pairs:
+            for lang_pair in self.lang_pairs:
                 _, tgt = lang_pair.split('-')
                 sample_key = _get_denoising_dataset_key(lang_pair)
                 forward_backward(model.models['{0}-{0}'.format(tgt)], sample[sample_key], sample_key, self.lambda_denoising)
@@ -395,12 +395,12 @@ class SemisupervisedTranslationTask(MultilingualTranslationTask):
             for logging_output in logging_outputs
             for key in logging_output
         }
-        lang_pair_keys = set(self.args.lang_pairs + [
+        lang_pair_keys = set(self.lang_pairs + [
             _get_bt_dataset_key(lang_pair)
-            for lang_pair in self.args.lang_pairs
+            for lang_pair in self.lang_pairs
         ] + [
             _get_denoising_dataset_key(lang_pair)
-            for lang_pair in self.args.lang_pairs
+            for lang_pair in self.lang_pairs
         ])
         logging_output_keys = logging_output_keys.intersection(lang_pair_keys)
         return super().aggregate_logging_outputs(logging_outputs, criterion, logging_output_keys)