Make tests work again (#4865)

* New import paths

* Duplicate entries

* Dataset readers can't be lazy anymore

allennlp/data/dataset_readers/sharded_dataset_reader.py

@@ -30,23 +30,13 @@ class ShardedDatasetReader(DatasetReader):
 
     Registered as a `DatasetReader` with name "sharded".
 
-    This class accepts all additional parameters of any `DatasetReader` class via `**kwargs`.
-    We give priority to the values set in the constructor for the instance of this class.
-    Optionally, we will automatically inherit attributes from the `base_reader` when required.
-
     # Parameters
 
     base_reader : `DatasetReader`
         Reader with a read method that accepts a single file.
     """
 
     def __init__(self, base_reader: DatasetReader, **kwargs) -> None:
-        # ShardedDatasetReader is a wrapper for the original base_reader so some of the parameters like 'lazy'
-        # can be safely inherited. However, ShardedDatasetReader is a class instance of a DatasetReader as well.
-        # So we give priority to the parameters for the current instance stored in 'kwargs'.
-        # If not present, we check the ones in the base reader
-        kwargs["lazy"] = kwargs.get("lazy", base_reader.lazy)
-
         super().__init__(
             manual_distributed_sharding=True, manual_multi_process_sharding=True, **kwargs
         )

allennlp/models/vision_text_model.py

@@ -5,6 +5,7 @@
 from overrides import overrides
 import numpy as np
 import torch
+from transformers import AutoModel
 
 from allennlp.data.fields.text_field import TextFieldTensors
 from allennlp.data.vocabulary import Vocabulary
@@ -16,8 +17,6 @@
     TransformerPooler,
 )
 
-from transformers.modeling_auto import AutoModel
-
 logger = logging.getLogger(__name__)
 
 
@@ -107,7 +106,7 @@ def from_huggingface_model_name(
         if hasattr(transformer.config, "embedding_size"):
             config = transformer.config
 
-            from transformers.modeling_albert import AlbertModel
+            from transformers.models.albert.modeling_albert import AlbertModel
 
             if isinstance(transformer, AlbertModel):
                 linear_transform = deepcopy(transformer.encoder.embedding_hidden_mapping_in)

allennlp/modules/elmo.py

@@ -13,7 +13,7 @@
 from allennlp.common.file_utils import cached_path
 from allennlp.common.util import lazy_groups_of
 from allennlp.data.instance import Instance
-from allennlp.data.tokenizers.token import Token
+from allennlp.data.tokenizers.token_class import Token
 from allennlp.data.vocabulary import Vocabulary
 from allennlp.data.batch import Batch
 from allennlp.data.fields import TextField

allennlp/modules/transformer/activation_layer.py

@@ -5,7 +5,7 @@
 
 from allennlp.modules.transformer.transformer_module import TransformerModule
 
-from transformers.modeling_bert import ACT2FN
+from transformers.models.bert.modeling_bert import ACT2FN
 
 
 class ActivationLayer(TransformerModule, FromParams):

allennlp/nn/activations.py

@@ -99,5 +99,4 @@ def _get_name(self):
     "softsign": (torch.nn.Softsign, None),
     "tanhshrink": (torch.nn.Tanhshrink, None),
     "selu": (torch.nn.SELU, None),
-    "gelu": (torch.nn.GELU, None),
 }

tests/data/dataset_readers/sharded_dataset_reader_test.py

@@ -73,20 +73,3 @@ def test_sharded_read_glob(self):
 
     def test_sharded_read_archive(self):
         self.read_and_check_instances(str(self.archive_filename))
-
-    def test_attributes_inheritance(self):
-        # current reader has lazy set to true
-        base_reader = SequenceTaggingDatasetReader(lazy=True)
-        reader = ShardedDatasetReader(base_reader=base_reader)
-
-        assert (
-            reader.lazy
-        ), "The ShardedDatasetReader didn't inherit the 'lazy' attribute from base_reader"
-
-    def test_set_attributes_main(self):
-        base_reader = SequenceTaggingDatasetReader(lazy=True)
-        reader = ShardedDatasetReader(base_reader=base_reader, lazy=False)
-
-        assert (
-            not reader.lazy
-        ), "The ShardedDatasetReader inherited the 'lazy' attribute from base_reader. It should be False"

tests/models/vilbert_vqa_test.py

@@ -1,4 +1,4 @@
-from transformers.modeling_auto import AutoModel
+from transformers import AutoModel
 
 from allennlp.common.testing import ModelTestCase
 from allennlp.data import Vocabulary

tests/models/visual_entailment_test.py

@@ -1,4 +1,4 @@
-from transformers.modeling_auto import AutoModel
+from transformers import AutoModel
 
 from allennlp.common.testing import ModelTestCase
 from allennlp.data import Vocabulary

tests/modules/transformer/self_attention_test.py

@@ -9,14 +9,14 @@
 from allennlp.modules.transformer import SelfAttention
 from allennlp.common.testing import AllenNlpTestCase
 
-from transformers.configuration_bert import BertConfig
-from transformers.modeling_bert import BertSelfAttention
-from transformers.configuration_roberta import RobertaConfig
-from transformers.modeling_roberta import RobertaSelfAttention
-from transformers.configuration_electra import ElectraConfig
-from transformers.modeling_electra import ElectraSelfAttention
-from transformers.configuration_distilbert import DistilBertConfig
-from transformers.modeling_distilbert import MultiHeadSelfAttention
+from transformers.models.bert.configuration_bert import BertConfig
+from transformers.models.bert.modeling_bert import BertSelfAttention
+from transformers.models.roberta.configuration_roberta import RobertaConfig
+from transformers.models.roberta.modeling_roberta import RobertaSelfAttention
+from transformers.models.electra.configuration_electra import ElectraConfig
+from transformers.models.electra.modeling_electra import ElectraSelfAttention
+from transformers.models.distilbert.configuration_distilbert import DistilBertConfig
+from transformers.models.distilbert.modeling_distilbert import MultiHeadSelfAttention
 
 PARAMS_DICT = {
     "hidden_size": 6,

tests/modules/transformer/toolkit_test.py

@@ -1,7 +1,6 @@
 import torch
 from overrides import overrides
-
-from transformers.modeling_albert import AlbertEmbeddings
+from transformers.models.albert.modeling_albert import AlbertEmbeddings
 
 from allennlp.common import cached_transformers
 from allennlp.common.testing import assert_equal_parameters

tests/modules/transformer/transformer_block_test.py

@@ -9,12 +9,12 @@
 from allennlp.modules.transformer import TransformerBlock
 from allennlp.common.testing import AllenNlpTestCase
 
-from transformers.configuration_bert import BertConfig
-from transformers.modeling_bert import BertEncoder
-from transformers.configuration_roberta import RobertaConfig
-from transformers.modeling_roberta import RobertaEncoder
-from transformers.configuration_electra import ElectraConfig
-from transformers.modeling_electra import ElectraEncoder
+from transformers.models.bert.configuration_bert import BertConfig
+from transformers.models.bert.modeling_bert import BertEncoder
+from transformers.models.roberta.configuration_roberta import RobertaConfig
+from transformers.models.roberta.modeling_roberta import RobertaEncoder
+from transformers.models.electra.configuration_electra import ElectraConfig
+from transformers.models.electra.modeling_electra import ElectraEncoder
 
 PARAMS_DICT = {
     "num_hidden_layers": 3,

tests/modules/transformer/transformer_layer_test.py

@@ -8,12 +8,12 @@
 from allennlp.modules.transformer import AttentionLayer, TransformerLayer
 from allennlp.common.testing import AllenNlpTestCase
 
-from transformers.configuration_bert import BertConfig
-from transformers.modeling_bert import BertAttention, BertLayer
-from transformers.configuration_roberta import RobertaConfig
-from transformers.modeling_roberta import RobertaAttention, RobertaLayer
-from transformers.configuration_electra import ElectraConfig
-from transformers.modeling_electra import ElectraAttention, ElectraLayer
+from transformers.models.bert.configuration_bert import BertConfig
+from transformers.models.bert.modeling_bert import BertAttention, BertLayer
+from transformers.models.roberta.configuration_roberta import RobertaConfig
+from transformers.models.roberta.modeling_roberta import RobertaAttention, RobertaLayer
+from transformers.models.electra.configuration_electra import ElectraConfig
+from transformers.models.electra.modeling_electra import ElectraAttention, ElectraLayer
 
 ATTENTION_PARAMS_DICT = {
     "hidden_size": 6,