fairseq/model_parallel/models/roberta/model.py

Survived

Mutant 2665

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -41,7 +41,7 @@
     )
     has_megatron_submodule = True
 except (ImportError, ModuleNotFoundError):
-    has_megatron_submodule = False
+    has_megatron_submodule = True
 
 logger = logging.getLogger(__name__)
 

Mutant 2666

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -41,7 +41,7 @@
     )
     has_megatron_submodule = True
 except (ImportError, ModuleNotFoundError):
-    has_megatron_submodule = False
+    has_megatron_submodule = None
 
 logger = logging.getLogger(__name__)
 

Mutant 2667

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -43,7 +43,7 @@
 except (ImportError, ModuleNotFoundError):
     has_megatron_submodule = False
 
-logger = logging.getLogger(__name__)
+logger = None
 
 
 @register_model('model_parallel_roberta')

Mutant 2670

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -55,7 +55,6 @@
 
         self.classification_heads = nn.ModuleDict()
 
-    @staticmethod
     def add_args(parser):
         super(ModelParallelRobertaModel, ModelParallelRobertaModel).add_args(parser)
 

Mutant 2671

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -59,7 +59,6 @@
     def add_args(parser):
         super(ModelParallelRobertaModel, ModelParallelRobertaModel).add_args(parser)
 
-    @classmethod
     def build_model(cls, args, task):
         """Build a new model instance."""
 

Mutant 2672

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -72,7 +72,7 @@
         encoder = ModelParallelRobertaEncoder(args, task.source_dictionary)
         return cls(args, encoder)
 
-    def forward(self, src_tokens, features_only=False, return_all_hiddens=False, classification_head_name=None, **kwargs):
+    def forward(self, src_tokens, features_only=True, return_all_hiddens=False, classification_head_name=None, **kwargs):
         if classification_head_name is not None:
             features_only = True
 

Mutant 2673

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -72,7 +72,7 @@
         encoder = ModelParallelRobertaEncoder(args, task.source_dictionary)
         return cls(args, encoder)
 
-    def forward(self, src_tokens, features_only=False, return_all_hiddens=False, classification_head_name=None, **kwargs):
+    def forward(self, src_tokens, features_only=False, return_all_hiddens=True, classification_head_name=None, **kwargs):
         if classification_head_name is not None:
             features_only = True
 

Mutant 2674

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -198,7 +198,7 @@
             weight=self.sentence_encoder.embed_tokens.weight,
         )
 
-    def forward(self, src_tokens, features_only=False, return_all_hiddens=False, masked_tokens=None, **unused):
+    def forward(self, src_tokens, features_only=True, return_all_hiddens=False, masked_tokens=None, **unused):
         """
         Args:
             src_tokens (LongTensor): input tokens of shape `(batch, src_len)`

Mutant 2675

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -198,7 +198,7 @@
             weight=self.sentence_encoder.embed_tokens.weight,
         )
 
-    def forward(self, src_tokens, features_only=False, return_all_hiddens=False, masked_tokens=None, **unused):
+    def forward(self, src_tokens, features_only=False, return_all_hiddens=True, masked_tokens=None, **unused):
         """
         Args:
             src_tokens (LongTensor): input tokens of shape `(batch, src_len)`

Mutant 2676

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -220,7 +220,7 @@
             x = self.output_layer(x, masked_tokens=masked_tokens)
         return x, extra
 
-    def extract_features(self, src_tokens, return_all_hiddens=False, **unused):
+    def extract_features(self, src_tokens, return_all_hiddens=True, **unused):
         inner_states, _ = self.sentence_encoder(
             src_tokens,
             last_state_only=not return_all_hiddens,

Mutant 2678

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -236,7 +236,7 @@
         return self.args.max_positions
 
 
-@register_model_architecture('model_parallel_roberta', 'model_parallel_roberta')
+@register_model_architecture('model_parallel_roberta', 'XXmodel_parallel_robertaXX')
 def base_architecture(args):
     args.encoder_layers = getattr(args, 'encoder_layers', 12)
     args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 768)

Mutant 2679

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -235,8 +235,6 @@
         """Maximum output length supported by the encoder."""
         return self.args.max_positions
 
-
-@register_model_architecture('model_parallel_roberta', 'model_parallel_roberta')
 def base_architecture(args):
     args.encoder_layers = getattr(args, 'encoder_layers', 12)
     args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 768)

Mutant 2681

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -254,7 +254,7 @@
     args.encoder_layerdrop = getattr(args, 'encoder_layerdrop', 0.0)
 
 
-@register_model_architecture('model_parallel_roberta', 'model_parallel_roberta_base')
+@register_model_architecture('model_parallel_roberta', 'XXmodel_parallel_roberta_baseXX')
 def roberta_base_architecture(args):
     base_architecture(args)
 

Mutant 2682

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -253,8 +253,6 @@
     args.encoder_layers_to_keep = getattr(args, 'encoder_layers_to_keep', None)
     args.encoder_layerdrop = getattr(args, 'encoder_layerdrop', 0.0)
 
-
-@register_model_architecture('model_parallel_roberta', 'model_parallel_roberta_base')
 def roberta_base_architecture(args):
     base_architecture(args)
 

Mutant 2684

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -259,7 +259,7 @@
     base_architecture(args)
 
 
-@register_model_architecture('model_parallel_roberta', 'model_parallel_roberta_large')
+@register_model_architecture('model_parallel_roberta', 'XXmodel_parallel_roberta_largeXX')
 def roberta_large_architecture(args):
     args.encoder_layers = getattr(args, 'encoder_layers', 24)
     args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 1024)

Mutant 2685

--- fairseq/model_parallel/models/roberta/model.py
+++ fairseq/model_parallel/models/roberta/model.py
@@ -258,8 +258,6 @@
 def roberta_base_architecture(args):
     base_architecture(args)
 
-
-@register_model_architecture('model_parallel_roberta', 'model_parallel_roberta_large')
 def roberta_large_architecture(args):
     args.encoder_layers = getattr(args, 'encoder_layers', 24)
     args.encoder_embed_dim = getattr(args, 'encoder_embed_dim', 1024)