fairseq/tasks/masked_lm.py

Killed 2 out of 48 mutants

Survived

Survived mutation testing. These mutants show holes in your test suite.

Mutant 2417

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -27,7 +27,7 @@
 from fairseq import utils
 
 
-logger = logging.getLogger(__name__)
+logger = None
 
 
 @register_task('masked_lm')

Mutant 2418

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -30,7 +30,7 @@
 logger = logging.getLogger(__name__)
 
 
-@register_task('masked_lm')
+@register_task('XXmasked_lmXX')
 class MaskedLMTask(FairseqTask):
     """Task for training masked language models (e.g., BERT, RoBERTa)."""
 

Mutant 2419

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -29,8 +29,6 @@
 
 logger = logging.getLogger(__name__)
 
-
-@register_task('masked_lm')
 class MaskedLMTask(FairseqTask):
     """Task for training masked language models (e.g., BERT, RoBERTa)."""
 

Mutant 2420

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -34,7 +34,6 @@
 class MaskedLMTask(FairseqTask):
     """Task for training masked language models (e.g., BERT, RoBERTa)."""
 
-    @staticmethod
     def add_args(parser):
         """Add task-specific arguments to the parser."""
         parser.add_argument('data', help='colon separated path to data directories list, \

Mutant 2421

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -37,7 +37,7 @@
     @staticmethod
     def add_args(parser):
         """Add task-specific arguments to the parser."""
-        parser.add_argument('data', help='colon separated path to data directories list, \
+        parser.add_argument('XXdataXX', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
         parser.add_argument('--sample-break-mode', default='complete',
                             choices=['none', 'complete', 'complete_doc', 'eos'],

Mutant 2422

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -37,8 +37,8 @@
     @staticmethod
     def add_args(parser):
         """Add task-specific arguments to the parser."""
-        parser.add_argument('data', help='colon separated path to data directories list, \
-                            will be iterated upon during epochs in round-robin manner')
+        parser.add_argument('data', help='XXcolon separated path to data directories list, \
+                            will be iterated upon during epochs in round-robin mannerXX')
         parser.add_argument('--sample-break-mode', default='complete',
                             choices=['none', 'complete', 'complete_doc', 'eos'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '

Mutant 2423

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -39,7 +39,7 @@
         """Add task-specific arguments to the parser."""
         parser.add_argument('data', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
-        parser.add_argument('--sample-break-mode', default='complete',
+        parser.add_argument('XX--sample-break-modeXX', default='complete',
                             choices=['none', 'complete', 'complete_doc', 'eos'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '
                                  'tokens. If set to "complete", splits samples only at the end '

Mutant 2424

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -39,7 +39,7 @@
         """Add task-specific arguments to the parser."""
         parser.add_argument('data', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
-        parser.add_argument('--sample-break-mode', default='complete',
+        parser.add_argument('--sample-break-mode', default='XXcompleteXX',
                             choices=['none', 'complete', 'complete_doc', 'eos'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '
                                  'tokens. If set to "complete", splits samples only at the end '

Mutant 2425

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -40,7 +40,7 @@
         parser.add_argument('data', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
         parser.add_argument('--sample-break-mode', default='complete',
-                            choices=['none', 'complete', 'complete_doc', 'eos'],
+                            choices=['XXnoneXX', 'complete', 'complete_doc', 'eos'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '
                                  'tokens. If set to "complete", splits samples only at the end '
                                  'of sentence, but may include multiple sentences per sample. '

Mutant 2426

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -40,7 +40,7 @@
         parser.add_argument('data', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
         parser.add_argument('--sample-break-mode', default='complete',
-                            choices=['none', 'complete', 'complete_doc', 'eos'],
+                            choices=['none', 'XXcompleteXX', 'complete_doc', 'eos'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '
                                  'tokens. If set to "complete", splits samples only at the end '
                                  'of sentence, but may include multiple sentences per sample. '

Mutant 2427

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -40,7 +40,7 @@
         parser.add_argument('data', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
         parser.add_argument('--sample-break-mode', default='complete',
-                            choices=['none', 'complete', 'complete_doc', 'eos'],
+                            choices=['none', 'complete', 'XXcomplete_docXX', 'eos'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '
                                  'tokens. If set to "complete", splits samples only at the end '
                                  'of sentence, but may include multiple sentences per sample. '

Mutant 2428

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -40,7 +40,7 @@
         parser.add_argument('data', help='colon separated path to data directories list, \
                             will be iterated upon during epochs in round-robin manner')
         parser.add_argument('--sample-break-mode', default='complete',
-                            choices=['none', 'complete', 'complete_doc', 'eos'],
+                            choices=['none', 'complete', 'complete_doc', 'XXeosXX'],
                             help='If omitted or "none", fills each sample with tokens-per-sample '
                                  'tokens. If set to "complete", splits samples only at the end '
                                  'of sentence, but may include multiple sentences per sample. '

Mutant 2429

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -41,7 +41,7 @@
                             will be iterated upon during epochs in round-robin manner')
         parser.add_argument('--sample-break-mode', default='complete',
                             choices=['none', 'complete', 'complete_doc', 'eos'],
-                            help='If omitted or "none", fills each sample with tokens-per-sample '
+                            help='XXIf omitted or "none", fills each sample with tokens-per-sample XX'
                                  'tokens. If set to "complete", splits samples only at the end '
                                  'of sentence, but may include multiple sentences per sample. '
                                  '"complete_doc" is similar but respects doc boundaries. '

Mutant 2430

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -46,7 +46,7 @@
                                  'of sentence, but may include multiple sentences per sample. '
                                  '"complete_doc" is similar but respects doc boundaries. '
                                  'If set to "eos", includes only one sentence per sample.')
-        parser.add_argument('--tokens-per-sample', default=512, type=int,
+        parser.add_argument('XX--tokens-per-sampleXX', default=512, type=int,
                             help='max number of total tokens over all segments '
                                  'per sample for BERT dataset')
         parser.add_argument('--mask-prob', default=0.15, type=float,

Mutant 2431

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -46,7 +46,7 @@
                                  'of sentence, but may include multiple sentences per sample. '
                                  '"complete_doc" is similar but respects doc boundaries. '
                                  'If set to "eos", includes only one sentence per sample.')
-        parser.add_argument('--tokens-per-sample', default=512, type=int,
+        parser.add_argument('--tokens-per-sample', default=513, type=int,
                             help='max number of total tokens over all segments '
                                  'per sample for BERT dataset')
         parser.add_argument('--mask-prob', default=0.15, type=float,

Mutant 2432

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -47,7 +47,7 @@
                                  '"complete_doc" is similar but respects doc boundaries. '
                                  'If set to "eos", includes only one sentence per sample.')
         parser.add_argument('--tokens-per-sample', default=512, type=int,
-                            help='max number of total tokens over all segments '
+                            help='XXmax number of total tokens over all segments XX'
                                  'per sample for BERT dataset')
         parser.add_argument('--mask-prob', default=0.15, type=float,
                             help='probability of replacing a token with mask')

Mutant 2433

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -49,7 +49,7 @@
         parser.add_argument('--tokens-per-sample', default=512, type=int,
                             help='max number of total tokens over all segments '
                                  'per sample for BERT dataset')
-        parser.add_argument('--mask-prob', default=0.15, type=float,
+        parser.add_argument('XX--mask-probXX', default=0.15, type=float,
                             help='probability of replacing a token with mask')
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
                             help='probability that a masked token is unmasked')

Mutant 2434

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -49,7 +49,7 @@
         parser.add_argument('--tokens-per-sample', default=512, type=int,
                             help='max number of total tokens over all segments '
                                  'per sample for BERT dataset')
-        parser.add_argument('--mask-prob', default=0.15, type=float,
+        parser.add_argument('--mask-prob', default=1.15, type=float,
                             help='probability of replacing a token with mask')
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
                             help='probability that a masked token is unmasked')

Mutant 2435

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -50,7 +50,7 @@
                             help='max number of total tokens over all segments '
                                  'per sample for BERT dataset')
         parser.add_argument('--mask-prob', default=0.15, type=float,
-                            help='probability of replacing a token with mask')
+                            help='XXprobability of replacing a token with maskXX')
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
                             help='probability that a masked token is unmasked')
         parser.add_argument('--random-token-prob', default=0.1, type=float,

Mutant 2436

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -51,7 +51,7 @@
                                  'per sample for BERT dataset')
         parser.add_argument('--mask-prob', default=0.15, type=float,
                             help='probability of replacing a token with mask')
-        parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
+        parser.add_argument('XX--leave-unmasked-probXX', default=0.1, type=float,
                             help='probability that a masked token is unmasked')
         parser.add_argument('--random-token-prob', default=0.1, type=float,
                             help='probability of replacing a token with a random token')

Mutant 2437

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -51,7 +51,7 @@
                                  'per sample for BERT dataset')
         parser.add_argument('--mask-prob', default=0.15, type=float,
                             help='probability of replacing a token with mask')
-        parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
+        parser.add_argument('--leave-unmasked-prob', default=1.1, type=float,
                             help='probability that a masked token is unmasked')
         parser.add_argument('--random-token-prob', default=0.1, type=float,
                             help='probability of replacing a token with a random token')

Mutant 2438

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -52,7 +52,7 @@
         parser.add_argument('--mask-prob', default=0.15, type=float,
                             help='probability of replacing a token with mask')
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
-                            help='probability that a masked token is unmasked')
+                            help='XXprobability that a masked token is unmaskedXX')
         parser.add_argument('--random-token-prob', default=0.1, type=float,
                             help='probability of replacing a token with a random token')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',

Mutant 2439

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -53,7 +53,7 @@
                             help='probability of replacing a token with mask')
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
                             help='probability that a masked token is unmasked')
-        parser.add_argument('--random-token-prob', default=0.1, type=float,
+        parser.add_argument('XX--random-token-probXX', default=0.1, type=float,
                             help='probability of replacing a token with a random token')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')

Mutant 2440

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -53,7 +53,7 @@
                             help='probability of replacing a token with mask')
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
                             help='probability that a masked token is unmasked')
-        parser.add_argument('--random-token-prob', default=0.1, type=float,
+        parser.add_argument('--random-token-prob', default=1.1, type=float,
                             help='probability of replacing a token with a random token')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')

Mutant 2441

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -54,7 +54,7 @@
         parser.add_argument('--leave-unmasked-prob', default=0.1, type=float,
                             help='probability that a masked token is unmasked')
         parser.add_argument('--random-token-prob', default=0.1, type=float,
-                            help='probability of replacing a token with a random token')
+                            help='XXprobability of replacing a token with a random tokenXX')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',

Mutant 2442

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -55,7 +55,7 @@
                             help='probability that a masked token is unmasked')
         parser.add_argument('--random-token-prob', default=0.1, type=float,
                             help='probability of replacing a token with a random token')
-        parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
+        parser.add_argument('XX--freq-weighted-replacementXX', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')

Mutant 2443

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -55,7 +55,7 @@
                             help='probability that a masked token is unmasked')
         parser.add_argument('--random-token-prob', default=0.1, type=float,
                             help='probability of replacing a token with a random token')
-        parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
+        parser.add_argument('--freq-weighted-replacement', default=True, action='store_true',
                             help='sample random replacement words based on word frequencies')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')

Mutant 2445

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -56,7 +56,7 @@
         parser.add_argument('--random-token-prob', default=0.1, type=float,
                             help='probability of replacing a token with a random token')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
-                            help='sample random replacement words based on word frequencies')
+                            help='XXsample random replacement words based on word frequenciesXX')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',

Mutant 2446

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -57,7 +57,7 @@
                             help='probability of replacing a token with a random token')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')
-        parser.add_argument('--mask-whole-words', default=False, action='store_true',
+        parser.add_argument('XX--mask-whole-wordsXX', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',
                             choices=['none', 'truncate', 'random_crop'],

Mutant 2447

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -57,7 +57,7 @@
                             help='probability of replacing a token with a random token')
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')
-        parser.add_argument('--mask-whole-words', default=False, action='store_true',
+        parser.add_argument('--mask-whole-words', default=True, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',
                             choices=['none', 'truncate', 'random_crop'],

Mutant 2449

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -58,7 +58,7 @@
         parser.add_argument('--freq-weighted-replacement', default=False, action='store_true',
                             help='sample random replacement words based on word frequencies')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
-                            help='mask whole words; you may also want to set --bpe')
+                            help='XXmask whole words; you may also want to set --bpeXX')
         parser.add_argument('--shorten-method', default='none',
                             choices=['none', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')

Mutant 2450

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -59,7 +59,7 @@
                             help='sample random replacement words based on word frequencies')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
-        parser.add_argument('--shorten-method', default='none',
+        parser.add_argument('XX--shorten-methodXX', default='none',
                             choices=['none', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
         parser.add_argument('--shorten-data-split-list', default='',

Mutant 2451

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -59,7 +59,7 @@
                             help='sample random replacement words based on word frequencies')
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
-        parser.add_argument('--shorten-method', default='none',
+        parser.add_argument('--shorten-method', default='XXnoneXX',
                             choices=['none', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
         parser.add_argument('--shorten-data-split-list', default='',

Mutant 2452

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -60,7 +60,7 @@
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',
-                            choices=['none', 'truncate', 'random_crop'],
+                            choices=['XXnoneXX', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
         parser.add_argument('--shorten-data-split-list', default='',
                             help='comma-separated list of dataset splits to apply shortening to, '

Mutant 2453

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -60,7 +60,7 @@
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',
-                            choices=['none', 'truncate', 'random_crop'],
+                            choices=['none', 'XXtruncateXX', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
         parser.add_argument('--shorten-data-split-list', default='',
                             help='comma-separated list of dataset splits to apply shortening to, '

Mutant 2454

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -60,7 +60,7 @@
         parser.add_argument('--mask-whole-words', default=False, action='store_true',
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',
-                            choices=['none', 'truncate', 'random_crop'],
+                            choices=['none', 'truncate', 'XXrandom_cropXX'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
         parser.add_argument('--shorten-data-split-list', default='',
                             help='comma-separated list of dataset splits to apply shortening to, '

Mutant 2455

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -61,7 +61,7 @@
                             help='mask whole words; you may also want to set --bpe')
         parser.add_argument('--shorten-method', default='none',
                             choices=['none', 'truncate', 'random_crop'],
-                            help='if not none, shorten sequences that exceed --tokens-per-sample')
+                            help='XXif not none, shorten sequences that exceed --tokens-per-sampleXX')
         parser.add_argument('--shorten-data-split-list', default='',
                             help='comma-separated list of dataset splits to apply shortening to, '
                                  'e.g., "train,valid" (default: all dataset splits)')

Mutant 2456

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -62,7 +62,7 @@
         parser.add_argument('--shorten-method', default='none',
                             choices=['none', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
-        parser.add_argument('--shorten-data-split-list', default='',
+        parser.add_argument('XX--shorten-data-split-listXX', default='',
                             help='comma-separated list of dataset splits to apply shortening to, '
                                  'e.g., "train,valid" (default: all dataset splits)')
 

Mutant 2457

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -62,7 +62,7 @@
         parser.add_argument('--shorten-method', default='none',
                             choices=['none', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
-        parser.add_argument('--shorten-data-split-list', default='',
+        parser.add_argument('--shorten-data-split-list', default='XXXX',
                             help='comma-separated list of dataset splits to apply shortening to, '
                                  'e.g., "train,valid" (default: all dataset splits)')
 

Mutant 2458

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -63,7 +63,7 @@
                             choices=['none', 'truncate', 'random_crop'],
                             help='if not none, shorten sequences that exceed --tokens-per-sample')
         parser.add_argument('--shorten-data-split-list', default='',
-                            help='comma-separated list of dataset splits to apply shortening to, '
+                            help='XXcomma-separated list of dataset splits to apply shortening to, XX'
                                  'e.g., "train,valid" (default: all dataset splits)')
 
     def __init__(self, args, dictionary):

Mutant 2459

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -74,7 +74,6 @@
         # add mask token
         self.mask_idx = dictionary.add_symbol('')
 
-    @classmethod
     def setup_task(cls, args, **kwargs):
         paths = utils.split_paths(args.data)
         assert len(paths) > 0

Mutant 2460

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -82,7 +82,7 @@
         logger.info('dictionary: {} types'.format(len(dictionary)))
         return cls(args, dictionary)
 
-    def load_dataset(self, split, epoch=1, combine=False, **kwargs):
+    def load_dataset(self, split, epoch=2, combine=False, **kwargs):
         """Load a given dataset split.
 
         Args:

Mutant 2461

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -82,7 +82,7 @@
         logger.info('dictionary: {} types'.format(len(dictionary)))
         return cls(args, dictionary)
 
-    def load_dataset(self, split, epoch=1, combine=False, **kwargs):
+    def load_dataset(self, split, epoch=1, combine=True, **kwargs):
         """Load a given dataset split.
 
         Args:

Mutant 2462

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -173,7 +173,7 @@
             ],
         )
 
-    def build_dataset_for_inference(self, src_tokens, src_lengths, sort=True):
+    def build_dataset_for_inference(self, src_tokens, src_lengths, sort=False):
         src_dataset = PadDataset(
             TokenBlockDataset(
                 src_tokens,

Mutant 2463

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -201,7 +201,6 @@
             src_dataset = SortDataset(src_dataset, sort_order=[src_lengths])
         return src_dataset
 
-    @property
     def source_dictionary(self):
         return self.dictionary
 

Mutant 2464

--- fairseq/tasks/masked_lm.py
+++ fairseq/tasks/masked_lm.py
@@ -205,7 +205,6 @@
     def source_dictionary(self):
         return self.dictionary
 
-    @property
     def target_dictionary(self):
         return self.dictionary