[examples/language-modeling] Add dataset download instructions

Had to hunt for instructions to get the dataset used in this set of examples, so this PR proposes to add them to README.md.

@patrickvonplaten

examples/language-modeling/README.md

@@ -30,6 +30,15 @@ need extra processing on your datasets.
 The following examples, will run on a datasets hosted on our [hub](https://huggingface.co/datasets) or with your own
 text files for training and validation. We give examples of both below.
 
+### Datasets
+
+The main dataset used in the examples below is:
+
+```bash
+wget https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-raw-v1.zip
+unzip wikitext-2-raw-v1.zip
+```
+
 ### GPT-2/GPT and causal language modeling
 
 The following example fine-tunes GPT-2 on WikiText-2. We're using the raw WikiText-2 (no tokens were replaced before

src/transformers/models/bart/modeling_bart.py

@@ -110,12 +110,13 @@ def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int]
 
 
 def BartLayerNorm(normalized_shape: torch.Size, eps: float = 1e-5, elementwise_affine: bool = True):
-    try:
-        from apex.normalization import FusedLayerNorm
+    if torch.cuda.is_available():
+        try:
+            from apex.normalization import FusedLayerNorm
 
-        return FusedLayerNorm(normalized_shape, eps, elementwise_affine)
-    except ImportError:
-        pass
+            return FusedLayerNorm(normalized_shape, eps, elementwise_affine)
+        except ImportError:
+            pass
     return torch.nn.LayerNorm(normalized_shape, eps, elementwise_affine)
 
 

src/transformers/models/fsmt/modeling_fsmt.py

@@ -265,12 +265,14 @@ FSMT_INPUTS_DOCSTRING = r"""
 
 
 have_fused_layer_norm = False
-try:
-    from apex.normalization import FusedLayerNorm
+if torch.cuda.is_available():
+    try:
+        from apex.normalization import FusedLayerNorm
+
+        have_fused_layer_norm = True
+    except ImportError:
+        pass
 
-    have_fused_layer_norm = True
-except ImportError:
-    pass
 LayerNorm = FusedLayerNorm if have_fused_layer_norm else torch.nn.LayerNorm
 
 

src/transformers/models/prophetnet/modeling_prophetnet.py

@@ -511,12 +511,13 @@ class ProphetNetDecoderLMOutput(ModelOutput):
 
 
 def ProphetNetLayerNorm(normalized_shape, eps=1e-5, elementwise_affine=True):
-    try:
-        from apex.normalization import FusedLayerNorm
+    if torch.cuda.is_available():
+        try:
+            from apex.normalization import FusedLayerNorm
 
-        return FusedLayerNorm(normalized_shape, eps, elementwise_affine)
-    except ImportError:
-        pass
+            return FusedLayerNorm(normalized_shape, eps, elementwise_affine)
+        except ImportError:
+            pass
     return torch.nn.LayerNorm(normalized_shape, eps, elementwise_affine)