fairseq/models/masked_lm.py
Killed 12 out of 68 mutantsSurvived
Survived mutation testing. These mutants show holes in your test suite.Mutant 2178
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -24,7 +24,7 @@
from fairseq.modules.transformer_sentence_encoder import init_bert_params
-logger = logging.getLogger(__name__)
+logger = None
@register_model('masked_lm')
Mutant 2181
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -43,7 +43,6 @@
if getattr(args, 'apply_bert_init', False):
self.apply(init_bert_params)
- @staticmethod
def add_args(parser):
"""Add model-specific arguments to the parser."""
# Arguments related to dropout
Mutant 2182
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -47,7 +47,7 @@
def add_args(parser):
"""Add model-specific arguments to the parser."""
# Arguments related to dropout
- parser.add_argument('--dropout', type=float, metavar='D',
+ parser.add_argument('XX--dropoutXX', type=float, metavar='D',
help='dropout probability')
parser.add_argument('--attention-dropout', type=float,
metavar='D', help='dropout probability for'
Mutant 2183
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -47,7 +47,7 @@
def add_args(parser):
"""Add model-specific arguments to the parser."""
# Arguments related to dropout
- parser.add_argument('--dropout', type=float, metavar='D',
+ parser.add_argument('--dropout', type=float, metavar='XXDXX',
help='dropout probability')
parser.add_argument('--attention-dropout', type=float,
metavar='D', help='dropout probability for'
Mutant 2184
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -48,7 +48,7 @@
"""Add model-specific arguments to the parser."""
# Arguments related to dropout
parser.add_argument('--dropout', type=float, metavar='D',
- help='dropout probability')
+ help='XXdropout probabilityXX')
parser.add_argument('--attention-dropout', type=float,
metavar='D', help='dropout probability for'
' attention weights')
Mutant 2185
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -49,7 +49,7 @@
# Arguments related to dropout
parser.add_argument('--dropout', type=float, metavar='D',
help='dropout probability')
- parser.add_argument('--attention-dropout', type=float,
+ parser.add_argument('XX--attention-dropoutXX', type=float,
metavar='D', help='dropout probability for'
' attention weights')
parser.add_argument('--act-dropout', type=float,
Mutant 2186
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -50,7 +50,7 @@
parser.add_argument('--dropout', type=float, metavar='D',
help='dropout probability')
parser.add_argument('--attention-dropout', type=float,
- metavar='D', help='dropout probability for'
+ metavar='XXDXX', help='dropout probability for'
' attention weights')
parser.add_argument('--act-dropout', type=float,
metavar='D', help='dropout probability after'
Mutant 2187
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -50,7 +50,7 @@
parser.add_argument('--dropout', type=float, metavar='D',
help='dropout probability')
parser.add_argument('--attention-dropout', type=float,
- metavar='D', help='dropout probability for'
+ metavar='D', help='XXdropout probability forXX'
' attention weights')
parser.add_argument('--act-dropout', type=float,
metavar='D', help='dropout probability after'
Mutant 2188
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -52,7 +52,7 @@
parser.add_argument('--attention-dropout', type=float,
metavar='D', help='dropout probability for'
' attention weights')
- parser.add_argument('--act-dropout', type=float,
+ parser.add_argument('XX--act-dropoutXX', type=float,
metavar='D', help='dropout probability after'
' activation in FFN')
Mutant 2189
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -53,7 +53,7 @@
metavar='D', help='dropout probability for'
' attention weights')
parser.add_argument('--act-dropout', type=float,
- metavar='D', help='dropout probability after'
+ metavar='XXDXX', help='dropout probability after'
' activation in FFN')
# Arguments related to hidden states and self-attention
Mutant 2190
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -53,7 +53,7 @@
metavar='D', help='dropout probability for'
' attention weights')
parser.add_argument('--act-dropout', type=float,
- metavar='D', help='dropout probability after'
+ metavar='D', help='XXdropout probability afterXX'
' activation in FFN')
# Arguments related to hidden states and self-attention
Mutant 2191
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -57,7 +57,7 @@
' activation in FFN')
# Arguments related to hidden states and self-attention
- parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N',
+ parser.add_argument('XX--encoder-ffn-embed-dimXX', type=int, metavar='N',
help='encoder embedding dimension for FFN')
parser.add_argument('--encoder-layers', type=int, metavar='N',
help='num encoder layers')
Mutant 2192
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -57,7 +57,7 @@
' activation in FFN')
# Arguments related to hidden states and self-attention
- parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N',
+ parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='XXNXX',
help='encoder embedding dimension for FFN')
parser.add_argument('--encoder-layers', type=int, metavar='N',
help='num encoder layers')
Mutant 2193
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -58,7 +58,7 @@
# Arguments related to hidden states and self-attention
parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N',
- help='encoder embedding dimension for FFN')
+ help='XXencoder embedding dimension for FFNXX')
parser.add_argument('--encoder-layers', type=int, metavar='N',
help='num encoder layers')
parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
Mutant 2194
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -59,7 +59,7 @@
# Arguments related to hidden states and self-attention
parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N',
help='encoder embedding dimension for FFN')
- parser.add_argument('--encoder-layers', type=int, metavar='N',
+ parser.add_argument('XX--encoder-layersXX', type=int, metavar='N',
help='num encoder layers')
parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
help='num encoder attention heads')
Mutant 2195
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -59,7 +59,7 @@
# Arguments related to hidden states and self-attention
parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N',
help='encoder embedding dimension for FFN')
- parser.add_argument('--encoder-layers', type=int, metavar='N',
+ parser.add_argument('--encoder-layers', type=int, metavar='XXNXX',
help='num encoder layers')
parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
help='num encoder attention heads')
Mutant 2196
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -60,7 +60,7 @@
parser.add_argument('--encoder-ffn-embed-dim', type=int, metavar='N',
help='encoder embedding dimension for FFN')
parser.add_argument('--encoder-layers', type=int, metavar='N',
- help='num encoder layers')
+ help='XXnum encoder layersXX')
parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
help='num encoder attention heads')
Mutant 2197
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -61,7 +61,7 @@
help='encoder embedding dimension for FFN')
parser.add_argument('--encoder-layers', type=int, metavar='N',
help='num encoder layers')
- parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
+ parser.add_argument('XX--encoder-attention-headsXX', type=int, metavar='N',
help='num encoder attention heads')
# Arguments related to input and output embeddings
Mutant 2198
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -61,7 +61,7 @@
help='encoder embedding dimension for FFN')
parser.add_argument('--encoder-layers', type=int, metavar='N',
help='num encoder layers')
- parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
+ parser.add_argument('--encoder-attention-heads', type=int, metavar='XXNXX',
help='num encoder attention heads')
# Arguments related to input and output embeddings
Mutant 2199
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -62,7 +62,7 @@
parser.add_argument('--encoder-layers', type=int, metavar='N',
help='num encoder layers')
parser.add_argument('--encoder-attention-heads', type=int, metavar='N',
- help='num encoder attention heads')
+ help='XXnum encoder attention headsXX')
# Arguments related to input and output embeddings
parser.add_argument('--encoder-embed-dim', type=int, metavar='N',
Mutant 2200
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -65,7 +65,7 @@
help='num encoder attention heads')
# Arguments related to input and output embeddings
- parser.add_argument('--encoder-embed-dim', type=int, metavar='N',
+ parser.add_argument('XX--encoder-embed-dimXX', type=int, metavar='N',
help='encoder embedding dimension')
parser.add_argument('--share-encoder-input-output-embed',
action='store_true', help='share encoder input'
Mutant 2201
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -65,7 +65,7 @@
help='num encoder attention heads')
# Arguments related to input and output embeddings
- parser.add_argument('--encoder-embed-dim', type=int, metavar='N',
+ parser.add_argument('--encoder-embed-dim', type=int, metavar='XXNXX',
help='encoder embedding dimension')
parser.add_argument('--share-encoder-input-output-embed',
action='store_true', help='share encoder input'
Mutant 2202
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -66,7 +66,7 @@
# Arguments related to input and output embeddings
parser.add_argument('--encoder-embed-dim', type=int, metavar='N',
- help='encoder embedding dimension')
+ help='XXencoder embedding dimensionXX')
parser.add_argument('--share-encoder-input-output-embed',
action='store_true', help='share encoder input'
' and output embeddings')
Mutant 2203
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -67,7 +67,7 @@
# Arguments related to input and output embeddings
parser.add_argument('--encoder-embed-dim', type=int, metavar='N',
help='encoder embedding dimension')
- parser.add_argument('--share-encoder-input-output-embed',
+ parser.add_argument('XX--share-encoder-input-output-embedXX',
action='store_true', help='share encoder input'
' and output embeddings')
parser.add_argument('--encoder-learned-pos', action='store_true',
Mutant 2205
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -68,7 +68,7 @@
parser.add_argument('--encoder-embed-dim', type=int, metavar='N',
help='encoder embedding dimension')
parser.add_argument('--share-encoder-input-output-embed',
- action='store_true', help='share encoder input'
+ action='store_true', help='XXshare encoder inputXX'
' and output embeddings')
parser.add_argument('--encoder-learned-pos', action='store_true',
help='use learned positional embeddings in the encoder')
Mutant 2206
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -70,7 +70,7 @@
parser.add_argument('--share-encoder-input-output-embed',
action='store_true', help='share encoder input'
' and output embeddings')
- parser.add_argument('--encoder-learned-pos', action='store_true',
+ parser.add_argument('XX--encoder-learned-posXX', action='store_true',
help='use learned positional embeddings in the encoder')
parser.add_argument('--no-token-positional-embeddings',
action='store_true',
Mutant 2208
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -71,7 +71,7 @@
action='store_true', help='share encoder input'
' and output embeddings')
parser.add_argument('--encoder-learned-pos', action='store_true',
- help='use learned positional embeddings in the encoder')
+ help='XXuse learned positional embeddings in the encoderXX')
parser.add_argument('--no-token-positional-embeddings',
action='store_true',
help='if set, disables positional embeddings'
Mutant 2209
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -72,7 +72,7 @@
' and output embeddings')
parser.add_argument('--encoder-learned-pos', action='store_true',
help='use learned positional embeddings in the encoder')
- parser.add_argument('--no-token-positional-embeddings',
+ parser.add_argument('XX--no-token-positional-embeddingsXX',
action='store_true',
help='if set, disables positional embeddings'
' (outside self attention)')
Mutant 2211
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -74,7 +74,7 @@
help='use learned positional embeddings in the encoder')
parser.add_argument('--no-token-positional-embeddings',
action='store_true',
- help='if set, disables positional embeddings'
+ help='XXif set, disables positional embeddingsXX'
' (outside self attention)')
parser.add_argument('--num-segment', type=int, metavar='N',
help='num segment in the input')
Mutant 2212
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -76,7 +76,7 @@
action='store_true',
help='if set, disables positional embeddings'
' (outside self attention)')
- parser.add_argument('--num-segment', type=int, metavar='N',
+ parser.add_argument('XX--num-segmentXX', type=int, metavar='N',
help='num segment in the input')
parser.add_argument('--max-positions', type=int,
help='number of positional embeddings to learn')
Mutant 2213
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -76,7 +76,7 @@
action='store_true',
help='if set, disables positional embeddings'
' (outside self attention)')
- parser.add_argument('--num-segment', type=int, metavar='N',
+ parser.add_argument('--num-segment', type=int, metavar='XXNXX',
help='num segment in the input')
parser.add_argument('--max-positions', type=int,
help='number of positional embeddings to learn')
Mutant 2214
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -77,7 +77,7 @@
help='if set, disables positional embeddings'
' (outside self attention)')
parser.add_argument('--num-segment', type=int, metavar='N',
- help='num segment in the input')
+ help='XXnum segment in the inputXX')
parser.add_argument('--max-positions', type=int,
help='number of positional embeddings to learn')
Mutant 2215
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -78,7 +78,7 @@
' (outside self attention)')
parser.add_argument('--num-segment', type=int, metavar='N',
help='num segment in the input')
- parser.add_argument('--max-positions', type=int,
+ parser.add_argument('XX--max-positionsXX', type=int,
help='number of positional embeddings to learn')
# Arguments related to sentence level prediction
Mutant 2216
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -79,7 +79,7 @@
parser.add_argument('--num-segment', type=int, metavar='N',
help='num segment in the input')
parser.add_argument('--max-positions', type=int,
- help='number of positional embeddings to learn')
+ help='XXnumber of positional embeddings to learnXX')
# Arguments related to sentence level prediction
parser.add_argument('--sentence-class-num', type=int, metavar='N',
Mutant 2217
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -82,7 +82,7 @@
help='number of positional embeddings to learn')
# Arguments related to sentence level prediction
- parser.add_argument('--sentence-class-num', type=int, metavar='N',
+ parser.add_argument('XX--sentence-class-numXX', type=int, metavar='N',
help='number of classes for sentence task')
parser.add_argument('--sent-loss', action='store_true', help='if set,'
' calculate sentence level predictions')
Mutant 2218
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -82,7 +82,7 @@
help='number of positional embeddings to learn')
# Arguments related to sentence level prediction
- parser.add_argument('--sentence-class-num', type=int, metavar='N',
+ parser.add_argument('--sentence-class-num', type=int, metavar='XXNXX',
help='number of classes for sentence task')
parser.add_argument('--sent-loss', action='store_true', help='if set,'
' calculate sentence level predictions')
Mutant 2219
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -83,7 +83,7 @@
# Arguments related to sentence level prediction
parser.add_argument('--sentence-class-num', type=int, metavar='N',
- help='number of classes for sentence task')
+ help='XXnumber of classes for sentence taskXX')
parser.add_argument('--sent-loss', action='store_true', help='if set,'
' calculate sentence level predictions')
Mutant 2220
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -84,7 +84,7 @@
# Arguments related to sentence level prediction
parser.add_argument('--sentence-class-num', type=int, metavar='N',
help='number of classes for sentence task')
- parser.add_argument('--sent-loss', action='store_true', help='if set,'
+ parser.add_argument('XX--sent-lossXX', action='store_true', help='if set,'
' calculate sentence level predictions')
# Arguments related to parameter initialization
Mutant 2222
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -84,7 +84,7 @@
# Arguments related to sentence level prediction
parser.add_argument('--sentence-class-num', type=int, metavar='N',
help='number of classes for sentence task')
- parser.add_argument('--sent-loss', action='store_true', help='if set,'
+ parser.add_argument('--sent-loss', action='store_true', help='XXif set,XX'
' calculate sentence level predictions')
# Arguments related to parameter initialization
Mutant 2223
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -88,7 +88,7 @@
' calculate sentence level predictions')
# Arguments related to parameter initialization
- parser.add_argument('--apply-bert-init', action='store_true',
+ parser.add_argument('XX--apply-bert-initXX', action='store_true',
help='use custom param initialization for BERT')
# misc params
Mutant 2225
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -89,7 +89,7 @@
# Arguments related to parameter initialization
parser.add_argument('--apply-bert-init', action='store_true',
- help='use custom param initialization for BERT')
+ help='XXuse custom param initialization for BERTXX')
# misc params
parser.add_argument('--activation-fn',
Mutant 2226
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -92,7 +92,7 @@
help='use custom param initialization for BERT')
# misc params
- parser.add_argument('--activation-fn',
+ parser.add_argument('XX--activation-fnXX',
choices=utils.get_available_activation_fns(),
help='activation function to use')
parser.add_argument('--pooler-activation-fn',
Mutant 2227
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -94,7 +94,7 @@
# misc params
parser.add_argument('--activation-fn',
choices=utils.get_available_activation_fns(),
- help='activation function to use')
+ help='XXactivation function to useXX')
parser.add_argument('--pooler-activation-fn',
choices=utils.get_available_activation_fns(),
help='Which activation function to use for pooler layer.')
Mutant 2228
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -95,7 +95,7 @@
parser.add_argument('--activation-fn',
choices=utils.get_available_activation_fns(),
help='activation function to use')
- parser.add_argument('--pooler-activation-fn',
+ parser.add_argument('XX--pooler-activation-fnXX',
choices=utils.get_available_activation_fns(),
help='Which activation function to use for pooler layer.')
parser.add_argument('--encoder-normalize-before', action='store_true',
Mutant 2229
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -97,7 +97,7 @@
help='activation function to use')
parser.add_argument('--pooler-activation-fn',
choices=utils.get_available_activation_fns(),
- help='Which activation function to use for pooler layer.')
+ help='XXWhich activation function to use for pooler layer.XX')
parser.add_argument('--encoder-normalize-before', action='store_true',
help='apply layernorm before each encoder block')
Mutant 2230
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -98,7 +98,7 @@
parser.add_argument('--pooler-activation-fn',
choices=utils.get_available_activation_fns(),
help='Which activation function to use for pooler layer.')
- parser.add_argument('--encoder-normalize-before', action='store_true',
+ parser.add_argument('XX--encoder-normalize-beforeXX', action='store_true',
help='apply layernorm before each encoder block')
def forward(self, src_tokens, segment_labels=None, **kwargs):
Mutant 2232
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -99,7 +99,7 @@
choices=utils.get_available_activation_fns(),
help='Which activation function to use for pooler layer.')
parser.add_argument('--encoder-normalize-before', action='store_true',
- help='apply layernorm before each encoder block')
+ help='XXapply layernorm before each encoder blockXX')
def forward(self, src_tokens, segment_labels=None, **kwargs):
return self.encoder(src_tokens, segment_labels=segment_labels, **kwargs)
Mutant 2233
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -107,7 +107,6 @@
def max_positions(self):
return self.encoder.max_positions
- @classmethod
def build_model(cls, args, task):
"""Build a new model instance."""
# make sure all arguments are present in older models
Mutant 2235
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -267,7 +267,7 @@
return state_dict
-@register_model_architecture('masked_lm', 'masked_lm')
+@register_model_architecture('masked_lm', 'XXmasked_lmXX')
def base_architecture(args):
args.dropout = getattr(args, 'dropout', 0.1)
args.attention_dropout = getattr(args, 'attention_dropout', 0.1)
Mutant 2236
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -266,8 +266,6 @@
del state_dict[k]
return state_dict
-
-@register_model_architecture('masked_lm', 'masked_lm')
def base_architecture(args):
args.dropout = getattr(args, 'dropout', 0.1)
args.attention_dropout = getattr(args, 'attention_dropout', 0.1)
Mutant 2238
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -293,7 +293,7 @@
args.encoder_normalize_before = getattr(args, 'encoder_normalize_before', False)
-@register_model_architecture('masked_lm', 'bert_base')
+@register_model_architecture('masked_lm', 'XXbert_baseXX')
def bert_base_architecture(args):
args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 768)
args.share_encoder_input_output_embed = getattr(
Mutant 2239
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -292,8 +292,6 @@
args.pooler_activation_fn = getattr(args, 'pooler_activation_fn', 'tanh')
args.encoder_normalize_before = getattr(args, 'encoder_normalize_before', False)
-
-@register_model_architecture('masked_lm', 'bert_base')
def bert_base_architecture(args):
args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 768)
args.share_encoder_input_output_embed = getattr(
Mutant 2241
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -319,7 +319,7 @@
base_architecture(args)
-@register_model_architecture('masked_lm', 'bert_large')
+@register_model_architecture('masked_lm', 'XXbert_largeXX')
def bert_large_architecture(args):
args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 1024)
args.encoder_layers = getattr(args, 'encoder_layers', 24)
Mutant 2242
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -318,8 +318,6 @@
args.encoder_normalize_before = getattr(args, 'encoder_normalize_before', True)
base_architecture(args)
-
-@register_model_architecture('masked_lm', 'bert_large')
def bert_large_architecture(args):
args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 1024)
args.encoder_layers = getattr(args, 'encoder_layers', 24)
Mutant 2244
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -328,7 +328,7 @@
bert_base_architecture(args)
-@register_model_architecture('masked_lm', 'xlm_base')
+@register_model_architecture('masked_lm', 'XXxlm_baseXX')
def xlm_architecture(args):
args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 1024)
args.share_encoder_input_output_embed = getattr(
Mutant 2245
--- fairseq/models/masked_lm.py
+++ fairseq/models/masked_lm.py
@@ -327,8 +327,6 @@
args.encoder_ffn_embed_dim = getattr(args, 'encoder_ffn_embed_dim', 4096)
bert_base_architecture(args)
-
-@register_model_architecture('masked_lm', 'xlm_base')
def xlm_architecture(args):
args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 1024)
args.share_encoder_input_output_embed = getattr(