[BART] generation_mode as a kwarg not a class attribute (#3278)

src/transformers/modeling_bart.py

@@ -437,7 +437,6 @@ class BartDecoder(nn.Module):
             [DecoderLayer(config) for _ in range(config.decoder_layers)]
         )  # type: List[DecoderLayer]
         self.layernorm_embedding = LayerNorm(config.d_model)
-        self.generation_mode = False
 
     def forward(
         self,
@@ -446,6 +445,7 @@ class BartDecoder(nn.Module):
         encoder_padding_mask,
         combined_mask,
         decoder_cached_states=None,
+        generation_mode=False,
         **unused
     ):
         """
@@ -474,9 +474,9 @@ class BartDecoder(nn.Module):
             assert encoder_padding_mask.max() <= 0
 
         # embed positions
-        positions = self.embed_positions(input_ids, generation_mode=self.generation_mode)
+        positions = self.embed_positions(input_ids, generation_mode=generation_mode)
 
-        if self.generation_mode:
+        if generation_mode:
             input_ids = input_ids[:, -1:]
             positions = positions[:, -1:]  # happens after we embed them
             assert input_ids.ne(self.padding_idx).any()
@@ -820,10 +820,11 @@ class BartModel(PretrainedBartModel):
         encoder_outputs=None,  # type: Tuple
         decoder_attention_mask=None,
         decoder_cached_states=None,
+        generation_mode=False,
     ):
 
         # make masks if user doesn't supply
-        if not self.decoder.generation_mode:
+        if not generation_mode:
             decoder_input_ids, decoder_attention_mask = _prepare_bart_decoder_inputs(
                 self.config,
                 input_ids,
@@ -842,6 +843,7 @@ class BartModel(PretrainedBartModel):
             attention_mask,
             decoder_attention_mask,
             decoder_cached_states=decoder_cached_states,
+            generation_mode=generation_mode,
         )
         # Attention and hidden_states will be [] or None if they aren't needed
         decoder_outputs = _filter_out_falsey_values(decoder_outputs)  # type: tuple
@@ -886,6 +888,7 @@ class BartForConditionalGeneration(PretrainedBartModel):
         decoder_attention_mask=None,
         decoder_cached_states=None,
         lm_labels=None,
+        generation_mode=False,
         **unused
     ):
         r"""
@@ -936,6 +939,7 @@ class BartForConditionalGeneration(PretrainedBartModel):
             encoder_outputs=encoder_outputs,
             decoder_attention_mask=decoder_attention_mask,
             decoder_cached_states=decoder_cached_states,
+            generation_mode=generation_mode,
         )
         lm_logits = self.lm_head(outputs[0])
         outputs = (lm_logits,) + outputs[1:]  # Add hidden states and attention if they are here
@@ -963,6 +967,7 @@ class BartForConditionalGeneration(PretrainedBartModel):
             "decoder_cached_states": decoder_cached_states,
             "decoder_input_ids": decoder_input_ids,
             "attention_mask": attention_mask,
+            "generation_mode": True,
         }
 
     def prepare_scores_for_generation(self, scores, cur_len, max_length):

src/transformers/modeling_utils.py

@@ -846,7 +846,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
             attention_mask = attention_mask.contiguous().view(
                 effective_batch_size * num_beams, input_ids_len
             )  # shape: (batch_size * num_return_sequences * num_beams, cur_len)
-
         if self.config.is_encoder_decoder:
             assert bos_token_id is not None, "Encoder Decoder Models need to have a bos_token_id"
             # encoder decoder need to start with empty input_ids and copy the input_ids to encoder_inputs
@@ -859,9 +858,6 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin):
             )
             cur_len = 1
 
-            # put model in generation mode if it has one
-            if hasattr(self.model, "decoder") and hasattr(self.model.decoder, "generation_mode"):
-                self.model.decoder.generation_mode = True
         else:
             encoder_inputs = None
             cur_len = input_ids.shape[-1]