try except sp

Author: DesmonDay

paddlenlp/transformers/__init__.py

@@ -29,16 +29,20 @@
 from .feature_extraction_utils import BatchFeature, FeatureExtractionMixin
 from .image_processing_utils import ImageProcessingMixin
 from .attention_utils import create_bigbird_rand_mask_idx_list
-from paddle.distributed.fleet.utils.sequence_parallel_utils import (
-    GatherOp,
-    ScatterOp,
-    AllGatherOp,
-    ReduceScatterOp,
-    ColumnSequenceParallelLinear,
-    RowSequenceParallelLinear,
-    mark_as_sequence_parallel_parameter,
-    register_sequence_parallel_allreduce_hooks,
-)
+
+try:
+    from paddle.distributed.fleet.utils.sequence_parallel_utils import (
+        GatherOp,
+        ScatterOp,
+        AllGatherOp,
+        ReduceScatterOp,
+        ColumnSequenceParallelLinear,
+        RowSequenceParallelLinear,
+        mark_as_sequence_parallel_parameter,
+        register_sequence_parallel_allreduce_hooks,
+    )
+except:
+    pass
 from .export import export_model
 
 # isort: split

paddlenlp/transformers/gpt/modeling.py

@@ -29,13 +29,17 @@
 from paddle.distributed import fleet
 from paddle.distributed.fleet.meta_parallel import get_rng_state_tracker
 from paddle.distributed.fleet.utils import recompute
-from paddle.distributed.fleet.utils.sequence_parallel_utils import (
-    ColumnSequenceParallelLinear,
-    GatherOp,
-    RowSequenceParallelLinear,
-    ScatterOp,
-    mark_as_sequence_parallel_parameter,
-)
+
+try:
+    from paddle.distributed.fleet.utils.sequence_parallel_utils import (
+        ColumnSequenceParallelLinear,
+        GatherOp,
+        RowSequenceParallelLinear,
+        ScatterOp,
+        mark_as_sequence_parallel_parameter,
+    )
+except:
+    pass
 from paddle.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
 
 from ...utils.converter import StateDictNameMapping

paddlenlp/transformers/gpt/modeling_pp.py

@@ -20,9 +20,13 @@
     SharedLayerDesc,
 )
 from paddle.distributed.fleet.utils import recompute
-from paddle.distributed.fleet.utils.sequence_parallel_utils import (
-    mark_as_sequence_parallel_parameter,
-)
+
+try:
+    from paddle.distributed.fleet.utils.sequence_parallel_utils import (
+        mark_as_sequence_parallel_parameter,
+    )
+except:
+    pass
 
 from paddlenlp.transformers.model_utils import PipelinePretrainedModel
 

paddlenlp/transformers/llama/modeling.py

@@ -44,13 +44,16 @@ def swiglu(x, y=None):
         return F.silu(x) * y
 
 
-from paddle.distributed.fleet.utils.sequence_parallel_utils import (
-    ColumnSequenceParallelLinear,
-    GatherOp,
-    RowSequenceParallelLinear,
-    ScatterOp,
-    mark_as_sequence_parallel_parameter,
-)
+try:
+    from paddle.distributed.fleet.utils.sequence_parallel_utils import (
+        ColumnSequenceParallelLinear,
+        GatherOp,
+        RowSequenceParallelLinear,
+        ScatterOp,
+        mark_as_sequence_parallel_parameter,
+    )
+except:
+    pass
 from paddle.utils import try_import
 
 from paddlenlp.transformers.conversion_utils import (

paddlenlp/transformers/mixtral/modeling.py

@@ -33,13 +33,16 @@
 except ImportError:
     fused_rotary_position_embedding = None
 
-from paddle.distributed.fleet.utils.sequence_parallel_utils import (
-    ColumnSequenceParallelLinear,
-    GatherOp,
-    RowSequenceParallelLinear,
-    ScatterOp,
-    mark_as_sequence_parallel_parameter,
-)
+try:
+    from paddle.distributed.fleet.utils.sequence_parallel_utils import (
+        ColumnSequenceParallelLinear,
+        GatherOp,
+        RowSequenceParallelLinear,
+        ScatterOp,
+        mark_as_sequence_parallel_parameter,
+    )
+except:
+    pass
 
 from paddlenlp.transformers.conversion_utils import (
     StateDictNameMapping,