make fix-copies

Author: steinmetzc

src/transformers/convert_slow_tokenizer.py

@@ -1563,8 +1563,6 @@ def bytes_to_unicode():
             n += 1
     cs = [chr(n) for n in cs]
     return dict(zip(bs, cs))
-
-
 class TikTokenConverter:
     """
     A general tiktoken converter.
@@ -1584,7 +1582,7 @@ def __init__(
         self.pattern = pattern
         self.add_prefix_space = add_prefix_space
         self.additional_special_tokens = (
-            additional_special_tokens.keys() if type(additional_special_tokens) is dict else additional_special_tokens
+            additional_special_tokens.keys() if isinstance(additional_special_tokens, dict) else additional_special_tokens
         )
 
     def extract_vocab_merges_from_model(self, tiktoken_url: str):

src/transformers/models/flaubert/modeling_flaubert.py

@@ -363,9 +363,9 @@ def forward(
         Returns:
             `torch.FloatTensor`: The end logits for SQuAD.
         """
-        assert start_states is not None or start_positions is not None, (
-            "One of start_states, start_positions should be not None"
-        )
+        assert (
+            start_states is not None or start_positions is not None
+        ), "One of start_states, start_positions should be not None"
         if start_positions is not None:
             slen, hsz = hidden_states.shape[-2:]
             start_positions = start_positions[:, None, None].expand(-1, -1, hsz)  # shape (bsz, 1, hsz)
@@ -432,9 +432,9 @@ def forward(
         """
         # No dependency on end_feature so that we can obtain one single `cls_logits` for each sample.
         hsz = hidden_states.shape[-1]
-        assert start_states is not None or start_positions is not None, (
-            "One of start_states, start_positions should be not None"
-        )
+        assert (
+            start_states is not None or start_positions is not None
+        ), "One of start_states, start_positions should be not None"
         if start_positions is not None:
             start_positions = start_positions[:, None, None].expand(-1, -1, hsz)  # shape (bsz, 1, hsz)
             start_states = hidden_states.gather(-2, start_positions).squeeze(-2)  # shape (bsz, hsz)

src/transformers/models/led/modeling_led.py

@@ -113,12 +113,12 @@ def __init__(self, config, layer_id):
 
         self.layer_id = layer_id
         attention_window = config.attention_window[self.layer_id]
-        assert attention_window % 2 == 0, (
-            f"`attention_window` for layer {self.layer_id} has to be an even value. Given {attention_window}"
-        )
-        assert attention_window > 0, (
-            f"`attention_window` for layer {self.layer_id} has to be positive. Given {attention_window}"
-        )
+        assert (
+            attention_window % 2 == 0
+        ), f"`attention_window` for layer {self.layer_id} has to be an even value. Given {attention_window}"
+        assert (
+            attention_window > 0
+        ), f"`attention_window` for layer {self.layer_id} has to be positive. Given {attention_window}"
 
         self.one_sided_attn_window_size = attention_window // 2
 
@@ -152,9 +152,9 @@ def forward(
         value_vectors = self.value(hidden_states)
 
         seq_len, batch_size, embed_dim = hidden_states.size()
-        assert embed_dim == self.embed_dim, (
-            f"hidden_states should have embed_dim = {self.embed_dim}, but has {embed_dim}"
-        )
+        assert (
+            embed_dim == self.embed_dim
+        ), f"hidden_states should have embed_dim = {self.embed_dim}, but has {embed_dim}"
 
         # normalize query
         query_vectors /= math.sqrt(self.head_dim)
@@ -222,9 +222,9 @@ def forward(
         )  # use fp32 for numerical stability
 
         if layer_head_mask is not None:
-            assert layer_head_mask.size() == (self.num_heads,), (
-                f"Head mask for a single layer should be of size {(self.num_heads,)}, but is {layer_head_mask.size()}"
-            )
+            assert layer_head_mask.size() == (
+                self.num_heads,
+            ), f"Head mask for a single layer should be of size {(self.num_heads,)}, but is {layer_head_mask.size()}"
             attn_probs = layer_head_mask.view(1, 1, -1, 1) * attn_probs
 
         # softmax sometimes inserts NaN if all positions are masked, replace them with 0
@@ -416,9 +416,9 @@ def _sliding_chunks_query_key_matmul(self, query: torch.Tensor, key: torch.Tenso
         overlap of size window_overlap
         """
         batch_size, seq_len, num_heads, head_dim = query.size()
-        assert seq_len % (window_overlap * 2) == 0, (
-            f"Sequence length should be multiple of {window_overlap * 2}. Given {seq_len}"
-        )
+        assert (
+            seq_len % (window_overlap * 2) == 0
+        ), f"Sequence length should be multiple of {window_overlap * 2}. Given {seq_len}"
         assert query.size() == key.size()
 
         chunks_count = torch.div(seq_len, window_overlap, rounding_mode="trunc") - 1
@@ -689,9 +689,9 @@ def _compute_global_attn_output_from_hidden(
 
         # apply layer head masking
         if layer_head_mask is not None:
-            assert layer_head_mask.size() == (self.num_heads,), (
-                f"Head mask for a single layer should be of size {(self.num_heads,)}, but is {layer_head_mask.size()}"
-            )
+            assert layer_head_mask.size() == (
+                self.num_heads,
+            ), f"Head mask for a single layer should be of size {(self.num_heads,)}, but is {layer_head_mask.size()}"
             global_attn_probs_float = layer_head_mask.view(1, -1, 1, 1) * global_attn_probs_float.view(
                 batch_size, self.num_heads, max_num_global_attn_indices, seq_len
             )

src/transformers/models/led/modeling_tf_led.py

@@ -182,12 +182,12 @@ def __init__(self, config, layer_id, **kwargs):
         self.layer_id = layer_id
         attention_window = config.attention_window[self.layer_id]
 
-        assert attention_window % 2 == 0, (
-            f"`attention_window` for layer {self.layer_id} has to be an even value. Given {attention_window}"
-        )
-        assert attention_window > 0, (
-            f"`attention_window` for layer {self.layer_id} has to be positive. Given {attention_window}"
-        )
+        assert (
+            attention_window % 2 == 0
+        ), f"`attention_window` for layer {self.layer_id} has to be an even value. Given {attention_window}"
+        assert (
+            attention_window > 0
+        ), f"`attention_window` for layer {self.layer_id} has to be positive. Given {attention_window}"
 
         self.one_sided_attn_window_size = attention_window // 2
 

src/transformers/models/xlnet/modeling_xlnet.py

@@ -605,9 +605,9 @@ def forward(
         Returns:
             `torch.FloatTensor`: The end logits for SQuAD.
         """
-        assert start_states is not None or start_positions is not None, (
-            "One of start_states, start_positions should be not None"
-        )
+        assert (
+            start_states is not None or start_positions is not None
+        ), "One of start_states, start_positions should be not None"
         if start_positions is not None:
             slen, hsz = hidden_states.shape[-2:]
             start_positions = start_positions[:, None, None].expand(-1, -1, hsz)  # shape (bsz, 1, hsz)
@@ -674,9 +674,9 @@ def forward(
         """
         # No dependency on end_feature so that we can obtain one single `cls_logits` for each sample.
         hsz = hidden_states.shape[-1]
-        assert start_states is not None or start_positions is not None, (
-            "One of start_states, start_positions should be not None"
-        )
+        assert (
+            start_states is not None or start_positions is not None
+        ), "One of start_states, start_positions should be not None"
         if start_positions is not None:
             start_positions = start_positions[:, None, None].expand(-1, -1, hsz)  # shape (bsz, 1, hsz)
             start_states = hidden_states.gather(-2, start_positions).squeeze(-2)  # shape (bsz, hsz)