Scatter dummies + skip pipeline tests (#13996)

* Scatter dummies + skip pipeline tests

* Add torch scatter to build docs

.circleci/config.yml

@@ -753,6 +753,7 @@ jobs:
             - run: sudo apt-get -y update && sudo apt-get install -y libsndfile1-dev
             - run: pip install --upgrade pip
             - run: pip install ."[docs]"
+            - run: pip install torch-scatter -f https://pytorch-geometric.com/whl/torch-1.9.0+cpu.html
             - save_cache:
                   key: v0.4-build_doc-{{ checksum "setup.py" }}
                   paths:

src/transformers/__init__.py

@@ -44,6 +44,7 @@ from . import dependency_versions_check
 from .file_utils import (
     _LazyModule,
     is_flax_available,
+    is_scatter_available,
     is_sentencepiece_available,
     is_speech_available,
     is_tf_available,
@@ -488,6 +489,25 @@ else:
         name for name in dir(dummy_timm_objects) if not name.startswith("_")
     ]
 
+if is_scatter_available():
+    _import_structure["models.tapas"].extend(
+        [
+            "TAPAS_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "TapasForMaskedLM",
+            "TapasForQuestionAnswering",
+            "TapasForSequenceClassification",
+            "TapasModel",
+            "TapasPreTrainedModel",
+            "load_tf_weights_in_tapas",
+        ]
+    )
+else:
+    from .utils import dummy_scatter_objects
+
+    _import_structure["utils.dummy_scatter_objects"] = [
+        name for name in dir(dummy_scatter_objects) if not name.startswith("_")
+    ]
+
 # PyTorch-backed objects
 if is_torch_available():
     _import_structure["benchmark.benchmark"] = ["PyTorchBenchmark"]
@@ -1157,17 +1177,6 @@ if is_torch_available():
             "load_tf_weights_in_t5",
         ]
     )
-    _import_structure["models.tapas"].extend(
-        [
-            "TAPAS_PRETRAINED_MODEL_ARCHIVE_LIST",
-            "TapasForMaskedLM",
-            "TapasForQuestionAnswering",
-            "TapasForSequenceClassification",
-            "TapasModel",
-            "TapasPreTrainedModel",
-            "load_tf_weights_in_tapas",
-        ]
-    )
     _import_structure["models.transfo_xl"].extend(
         [
             "TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -2282,6 +2291,19 @@ if TYPE_CHECKING:
     else:
         from .utils.dummy_timm_objects import *
 
+    if is_scatter_available():
+        from .models.tapas import (
+            TAPAS_PRETRAINED_MODEL_ARCHIVE_LIST,
+            TapasForMaskedLM,
+            TapasForQuestionAnswering,
+            TapasForSequenceClassification,
+            TapasModel,
+            TapasPreTrainedModel,
+            load_tf_weights_in_tapas,
+        )
+    else:
+        from .utils.dummy_scatter_objects import *
+
     if is_torch_available():
         # Benchmarks
         from .benchmark.benchmark import PyTorchBenchmark
@@ -2847,15 +2869,6 @@ if TYPE_CHECKING:
             T5PreTrainedModel,
             load_tf_weights_in_t5,
         )
-        from .models.tapas import (
-            TAPAS_PRETRAINED_MODEL_ARCHIVE_LIST,
-            TapasForMaskedLM,
-            TapasForQuestionAnswering,
-            TapasForSequenceClassification,
-            TapasModel,
-            TapasPreTrainedModel,
-            load_tf_weights_in_tapas,
-        )
         from .models.transfo_xl import (
             TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST,
             AdaptiveEmbedding,

src/transformers/utils/dummy_pt_objects.py

@@ -3487,58 +3487,6 @@ def load_tf_weights_in_t5(*args, **kwargs):
     requires_backends(load_tf_weights_in_t5, ["torch"])
 
 
-TAPAS_PRETRAINED_MODEL_ARCHIVE_LIST = None
-
-
-class TapasForMaskedLM:
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["torch"])
-
-    @classmethod
-    def from_pretrained(cls, *args, **kwargs):
-        requires_backends(cls, ["torch"])
-
-
-class TapasForQuestionAnswering:
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["torch"])
-
-    @classmethod
-    def from_pretrained(cls, *args, **kwargs):
-        requires_backends(cls, ["torch"])
-
-
-class TapasForSequenceClassification:
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["torch"])
-
-    @classmethod
-    def from_pretrained(cls, *args, **kwargs):
-        requires_backends(cls, ["torch"])
-
-
-class TapasModel:
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["torch"])
-
-    @classmethod
-    def from_pretrained(cls, *args, **kwargs):
-        requires_backends(cls, ["torch"])
-
-
-class TapasPreTrainedModel:
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["torch"])
-
-    @classmethod
-    def from_pretrained(cls, *args, **kwargs):
-        requires_backends(cls, ["torch"])
-
-
-def load_tf_weights_in_tapas(*args, **kwargs):
-    requires_backends(load_tf_weights_in_tapas, ["torch"])
-
-
 TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 

src/transformers/utils/dummy_scatter_objects.py

@@ -0,0 +1,54 @@
+# This file is autogenerated by the command `make fix-copies`, do not edit.
+from ..file_utils import requires_backends
+
+
+TAPAS_PRETRAINED_MODEL_ARCHIVE_LIST = None
+
+
+class TapasForMaskedLM:
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["scatter"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["scatter"])
+
+
+class TapasForQuestionAnswering:
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["scatter"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["scatter"])
+
+
+class TapasForSequenceClassification:
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["scatter"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["scatter"])
+
+
+class TapasModel:
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["scatter"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["scatter"])
+
+
+class TapasPreTrainedModel:
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["scatter"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["scatter"])
+
+
+def load_tf_weights_in_tapas(*args, **kwargs):
+    requires_backends(load_tf_weights_in_tapas, ["scatter"])

tests/test_pipelines_common.py

@@ -111,7 +111,12 @@ class PipelineTestCaseMeta(type):
                     tiny_config.is_encoder_decoder = False
                 if ModelClass.__name__.endswith("WithLMHead"):
                     tiny_config.is_decoder = True
-                model = ModelClass(tiny_config)
+                try:
+                    model = ModelClass(tiny_config)
+                except ImportError as e:
+                    self.skipTest(
+                        f"Cannot run with {tiny_config} as the model requires a library that isn't installed: {e}"
+                    )
                 if hasattr(model, "eval"):
                     model = model.eval()
                 if tokenizer_class is not None: