More PYUP fixes (#38883)

More pyup fixes

Signed-off-by: cyy <[email protected]>

Author: cyyever

src/transformers/generation/candidate_generator.py

@@ -710,8 +710,8 @@ def __init__(
         assistant_model: Optional["PreTrainedModel"] = None,
         assistant_prune_lm_head: bool = False,
     ):
-        self._target_tokenizer: "PreTrainedTokenizerBase" = target_tokenizer
-        self._assistant_tokenizer: "PreTrainedTokenizerBase" = assistant_tokenizer
+        self._target_tokenizer: PreTrainedTokenizerBase = target_tokenizer
+        self._assistant_tokenizer: PreTrainedTokenizerBase = assistant_tokenizer
         self._assistant_model_device: str = (
             assistant_model_device if assistant_model is None else assistant_model.device
         )

src/transformers/generation/streamers.py

@@ -72,7 +72,7 @@ class TextStreamer(BaseStreamer):
         ```
     """
 
-    def __init__(self, tokenizer: "AutoTokenizer", skip_prompt: bool = False, **decode_kwargs):
+    def __init__(self, tokenizer: AutoTokenizer, skip_prompt: bool = False, **decode_kwargs):
         self.tokenizer = tokenizer
         self.skip_prompt = skip_prompt
         self.decode_kwargs = decode_kwargs
@@ -206,7 +206,7 @@ class TextIteratorStreamer(TextStreamer):
     """
 
     def __init__(
-        self, tokenizer: "AutoTokenizer", skip_prompt: bool = False, timeout: Optional[float] = None, **decode_kwargs
+        self, tokenizer: AutoTokenizer, skip_prompt: bool = False, timeout: Optional[float] = None, **decode_kwargs
     ):
         super().__init__(tokenizer, skip_prompt, **decode_kwargs)
         self.text_queue = Queue()
@@ -284,7 +284,7 @@ class AsyncTextIteratorStreamer(TextStreamer):
     """
 
     def __init__(
-        self, tokenizer: "AutoTokenizer", skip_prompt: bool = False, timeout: Optional[float] = None, **decode_kwargs
+        self, tokenizer: AutoTokenizer, skip_prompt: bool = False, timeout: Optional[float] = None, **decode_kwargs
     ):
         super().__init__(tokenizer, skip_prompt, **decode_kwargs)
         self.text_queue = asyncio.Queue()

src/transformers/generation/utils.py

@@ -4723,7 +4723,7 @@ def _constrained_beam_search(
                 )
 
             if return_dict_in_generate and output_scores:
-                beam_indices = tuple((beam_indices[beam_idx[i]] + (beam_idx[i],) for i in range(len(beam_indices))))
+                beam_indices = tuple(beam_indices[beam_idx[i]] + (beam_idx[i],) for i in range(len(beam_indices)))
 
             # increase cur_len
             cur_len = cur_len + 1

src/transformers/integrations/integration_utils.py

@@ -1626,8 +1626,8 @@ def _log_model_checkpoint(self, source_directory: str, checkpoint: str):
                 target_path = consistent_checkpoint_path
             except OSError as e:
                 logger.warning(
-                    "NeptuneCallback was unable to made a copy of checkpoint due to I/O exception: '{}'. "
-                    "Could fail trying to upload.".format(e)
+                    f"NeptuneCallback was unable to made a copy of checkpoint due to I/O exception: '{e}'. "
+                    "Could fail trying to upload."
                 )
 
         self._metadata_namespace[self._target_checkpoints_namespace].upload_files(target_path)
@@ -1976,9 +1976,7 @@ def on_save(self, args, state, control, **kwargs):
                     )
                 except Exception as e:
                     logger.warning(
-                        "Could not remove checkpoint `{}` after going over the `save_total_limit`. Error is: {}".format(
-                            self._checkpoints_saved[0].name, e
-                        )
+                        f"Could not remove checkpoint `{self._checkpoints_saved[0].name}` after going over the `save_total_limit`. Error is: {e}"
                     )
                     break
                 self._checkpoints_saved = self._checkpoints_saved[1:]

src/transformers/modeling_tf_utils.py

@@ -1409,10 +1409,10 @@ def _save_checkpoint(self, checkpoint_dir, epoch):
 
     def prepare_tf_dataset(
         self,
-        dataset: "datasets.Dataset",  # noqa:F821
+        dataset: datasets.Dataset,  # noqa:F821
         batch_size: int = 8,
         shuffle: bool = True,
-        tokenizer: Optional["PreTrainedTokenizerBase"] = None,
+        tokenizer: Optional[PreTrainedTokenizerBase] = None,
         collate_fn: Optional[Callable] = None,
         collate_fn_args: Optional[dict[str, Any]] = None,
         drop_remainder: Optional[bool] = None,

src/transformers/modeling_utils.py

@@ -4424,10 +4424,8 @@ def from_pretrained(
                 raise ValueError("DeepSpeed Zero-3 is not compatible with passing a `device_map`.")
             if not is_accelerate_available():
                 raise ValueError(
-                    (
-                        "Using a `device_map`, `tp_plan`, `torch.device` context manager or setting `torch.set_default_device(device)` "
-                        "requires `accelerate`. You can install it with `pip install accelerate`"
-                    )
+                    "Using a `device_map`, `tp_plan`, `torch.device` context manager or setting `torch.set_default_device(device)` "
+                    "requires `accelerate`. You can install it with `pip install accelerate`"
                 )
 
         # handling bnb config from kwargs, remove after `load_in_{4/8}bit` deprecation.

src/transformers/models/albert/tokenization_albert.py

@@ -203,7 +203,7 @@ def _tokenize(self, text: str) -> list[str]:
         pieces = self.sp_model.encode(text, out_type=str)
         new_pieces = []
         for piece in pieces:
-            if len(piece) > 1 and piece[-1] == str(",") and piece[-2].isdigit():
+            if len(piece) > 1 and piece[-1] == "," and piece[-2].isdigit():
                 # Logic to handle special cases see https://github.com/google-research/bert/blob/master/README.md#tokenization
                 # `9,9` -> ['▁9', ',', '9'] instead of [`_9,`, '9']
                 cur_pieces = self.sp_model.EncodeAsPieces(piece[:-1].replace(SPIECE_UNDERLINE, ""))

src/transformers/models/bamba/modeling_bamba.py

@@ -830,7 +830,7 @@ def torch_forward(
 
             # 2. Compute the state for each intra-chunk
             # (right term of low-rank factorization of off-diagonal blocks; B terms)
-            decay_states = torch.exp((A_cumsum[:, :, :, -1:] - A_cumsum))
+            decay_states = torch.exp(A_cumsum[:, :, :, -1:] - A_cumsum)
             B_decay = B * decay_states.permute(0, -2, -1, 1)[..., None]
             states = (B_decay[..., None, :] * hidden_states[..., None]).sum(dim=2)
 

src/transformers/models/bamba/modular_bamba.py

@@ -632,7 +632,7 @@ def torch_forward(
 
             # 2. Compute the state for each intra-chunk
             # (right term of low-rank factorization of off-diagonal blocks; B terms)
-            decay_states = torch.exp((A_cumsum[:, :, :, -1:] - A_cumsum))
+            decay_states = torch.exp(A_cumsum[:, :, :, -1:] - A_cumsum)
             B_decay = B * decay_states.permute(0, -2, -1, 1)[..., None]
             states = (B_decay[..., None, :] * hidden_states[..., None]).sum(dim=2)
 

src/transformers/models/bart/tokenization_bart.py

@@ -32,7 +32,7 @@
 # See all BART models at https://huggingface.co/models?filter=bart
 
 
-@lru_cache()
+@lru_cache
 def bytes_to_unicode():
     """
     Returns list of utf-8 byte and a mapping to unicode strings. We specifically avoids mapping to whitespace/control

src/transformers/models/beit/modeling_beit.py

@@ -110,7 +110,7 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         return drop_path(hidden_states, self.drop_prob, self.training)
 
     def extra_repr(self) -> str:
-        return "p={}".format(self.drop_prob)
+        return f"p={self.drop_prob}"
 
 
 # Based on timm implementation, which can be found here:
@@ -513,8 +513,8 @@ def __init__(self, config: BeitConfig, window_size: Optional[tuple] = None, drop
 
         init_values = config.layer_scale_init_value
         if init_values > 0:
-            self.lambda_1 = nn.Parameter(init_values * torch.ones((config.hidden_size)), requires_grad=True)
-            self.lambda_2 = nn.Parameter(init_values * torch.ones((config.hidden_size)), requires_grad=True)
+            self.lambda_1 = nn.Parameter(init_values * torch.ones(config.hidden_size), requires_grad=True)
+            self.lambda_2 = nn.Parameter(init_values * torch.ones(config.hidden_size), requires_grad=True)
         else:
             self.lambda_1, self.lambda_2 = None, None
 

src/transformers/models/bert_japanese/tokenization_bert_japanese.py

@@ -934,7 +934,7 @@ def tokenize(self, text):
         pieces = self.sp_model.encode(text, out_type=str)
         new_pieces = []
         for piece in pieces:
-            if len(piece) > 1 and piece[-1] == str(",") and piece[-2].isdigit():
+            if len(piece) > 1 and piece[-1] == "," and piece[-2].isdigit():
                 cur_pieces = self.sp_model.EncodeAsPieces(piece[:-1].replace(SPIECE_UNDERLINE, ""))
                 if piece[0] != SPIECE_UNDERLINE and cur_pieces[0][0] == SPIECE_UNDERLINE:
                     if len(cur_pieces[0]) == 1:

src/transformers/models/biogpt/convert_biogpt_original_pytorch_checkpoint_to_pytorch.py

@@ -115,7 +115,7 @@ def add_from_file(self, f):
             except FileNotFoundError as fnfe:
                 raise fnfe
             except UnicodeError:
-                raise Exception("Incorrect encoding detected in {}, please rebuild the dataset".format(f))
+                raise Exception(f"Incorrect encoding detected in {f}, please rebuild the dataset")
             return
 
         lines = f.readlines()
@@ -133,11 +133,11 @@ def add_from_file(self, f):
                 word = line
                 if word in self and not overwrite:
                     raise RuntimeError(
-                        "Duplicate word found when loading Dictionary: '{}'. "
+                        f"Duplicate word found when loading Dictionary: '{word}'. "
                         "Duplicate words can overwrite earlier ones by adding the "
                         "#fairseq:overwrite flag at the end of the corresponding row "
                         "in the dictionary file. If using the Camembert model, please "
-                        "download an updated copy of the model file.".format(word)
+                        "download an updated copy of the model file."
                     )
                 self.add_symbol(word, n=count, overwrite=overwrite)
             except ValueError:

src/transformers/models/bit/modeling_bit.py

@@ -310,7 +310,7 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         return drop_path(hidden_states, self.drop_prob, self.training)
 
     def extra_repr(self) -> str:
-        return "p={}".format(self.drop_prob)
+        return f"p={self.drop_prob}"
 
 
 def make_div(value, divisor=8):

src/transformers/models/blenderbot/tokenization_blenderbot.py

@@ -35,7 +35,7 @@
 }
 
 
-@lru_cache()
+@lru_cache
 # Copied from transformers.models.roberta.tokenization_roberta.bytes_to_unicode
 def bytes_to_unicode():
     """

src/transformers/models/blip/modeling_blip_text.py

@@ -641,9 +641,7 @@ def get_extended_attention_mask(
                 extended_attention_mask = attention_mask[:, None, None, :]
         else:
             raise ValueError(
-                "Wrong shape for input_ids (shape {}) or attention_mask (shape {})".format(
-                    input_shape, attention_mask.shape
-                )
+                f"Wrong shape for input_ids (shape {input_shape}) or attention_mask (shape {attention_mask.shape})"
             )
 
         # Since attention_mask is 1.0 for positions we want to attend and 0.0 for
@@ -723,7 +721,7 @@ def forward(
         past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
 
         if attention_mask is None:
-            attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length))).to(device)
+            attention_mask = torch.ones((batch_size, seq_length + past_key_values_length)).to(device)
 
         # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
         # ourselves in which case we just need to make it broadcastable to all heads.

src/transformers/models/blip/modeling_tf_blip_text.py

@@ -800,9 +800,7 @@ def get_extended_attention_mask(
                 extended_attention_mask = attention_mask[:, None, None, :]
         else:
             raise ValueError(
-                "Wrong shape for input_ids (shape {}) or attention_mask (shape {})".format(
-                    input_shape, attention_mask.shape
-                )
+                f"Wrong shape for input_ids (shape {input_shape}) or attention_mask (shape {attention_mask.shape})"
             )
 
         # Since attention_mask is 1.0 for positions we want to attend and 0.0 for
@@ -881,7 +879,7 @@ def call(
         past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0
 
         if attention_mask is None:
-            attention_mask = tf.ones(((batch_size, seq_length + past_key_values_length)))
+            attention_mask = tf.ones((batch_size, seq_length + past_key_values_length))
 
         # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
         # ourselves in which case we just need to make it broadcastable to all heads.

src/transformers/models/blip_2/modeling_blip_2.py

@@ -1144,9 +1144,7 @@ def get_extended_attention_mask(
             extended_attention_mask = attention_mask[:, None, None, :]
         else:
             raise ValueError(
-                "Wrong shape for input_ids (shape {}) or attention_mask (shape {})".format(
-                    input_shape, attention_mask.shape
-                )
+                f"Wrong shape for input_ids (shape {input_shape}) or attention_mask (shape {attention_mask.shape})"
             )
 
         # Since attention_mask is 1.0 for positions we want to attend and 0.0 for

src/transformers/models/bloom/convert_bloom_original_checkpoint_to_pytorch.py

@@ -98,7 +98,7 @@ def convert_bloom_checkpoint_to_pytorch(
         config = BloomConfig()
 
         for j, file in enumerate(file_names):
-            print("Processing file: {}".format(file))
+            print(f"Processing file: {file}")
             tensors = None
 
             for i in range(pretraining_tp):
@@ -132,16 +132,16 @@ def convert_bloom_checkpoint_to_pytorch(
                 tensors,
                 os.path.join(
                     pytorch_dump_folder_path,
-                    "pytorch_model_{}-of-{}.bin".format(str(j + 1).zfill(5), str(len(file_names)).zfill(5)),
+                    f"pytorch_model_{str(j + 1).zfill(5)}-of-{str(len(file_names)).zfill(5)}.bin",
                 ),
             )
 
             for key in tensors.keys():
                 value = tensors[key]
                 total_size += value.numel() * get_dtype_size(value.dtype)
                 if key not in index_dict["weight_map"]:
-                    index_dict["weight_map"][key] = "pytorch_model_{}-of-{}.bin".format(
-                        str(j + 1).zfill(5), str(len(file_names)).zfill(5)
+                    index_dict["weight_map"][key] = (
+                        f"pytorch_model_{str(j + 1).zfill(5)}-of-{str(len(file_names)).zfill(5)}.bin"
                     )
 
         config = BloomConfig()

src/transformers/models/clap/modeling_clap.py

@@ -610,7 +610,7 @@ def get_attn_mask(self, height, width, dtype, device):
             mask_windows = window_partition(img_mask, self.window_size)
             mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
             attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
-            attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0))
+            attn_mask = attn_mask.masked_fill(attn_mask != 0, -100.0).masked_fill(attn_mask == 0, 0.0)
         else:
             attn_mask = None
         return attn_mask

src/transformers/models/clip/tokenization_clip.py

@@ -34,7 +34,7 @@
 }
 
 
-@lru_cache()
+@lru_cache
 def bytes_to_unicode():
     """
     Returns list of utf-8 byte and a mapping to unicode strings. We specifically avoids mapping to whitespace/control
@@ -488,7 +488,7 @@ def convert_tokens_to_string(self, tokens):
 
     def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> tuple[str]:
         if not os.path.isdir(save_directory):
-            logger.error("Vocabulary path ({}) should be a directory".format(save_directory))
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
             return
         vocab_file = os.path.join(
             save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
@@ -506,8 +506,8 @@ def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] =
             for bpe_tokens, token_index in sorted(self.bpe_ranks.items(), key=lambda kv: kv[1]):
                 if index != token_index:
                     logger.warning(
-                        "Saving vocabulary to {}: BPE merge indices are not consecutive."
-                        " Please check that the tokenizer is not corrupted!".format(merge_file)
+                        f"Saving vocabulary to {merge_file}: BPE merge indices are not consecutive."
+                        " Please check that the tokenizer is not corrupted!"
                     )
                     index = token_index
                 writer.write(" ".join(bpe_tokens) + "\n")

src/transformers/models/clipseg/convert_clipseg_original_pytorch_to_hf.py

@@ -181,7 +181,7 @@ def convert_clipseg_checkpoint(model_name, checkpoint_path, pytorch_dump_folder_
     missing_keys, unexpected_keys = model.load_state_dict(state_dict, strict=False)
 
     if missing_keys != ["clip.text_model.embeddings.position_ids", "clip.vision_model.embeddings.position_ids"]:
-        raise ValueError("Missing keys that are not expected: {}".format(missing_keys))
+        raise ValueError(f"Missing keys that are not expected: {missing_keys}")
     if unexpected_keys != ["decoder.reduce.weight", "decoder.reduce.bias"]:
         raise ValueError(f"Unexpected keys: {unexpected_keys}")
 

src/transformers/models/clvp/tokenization_clvp.py

@@ -34,7 +34,7 @@
 }
 
 
-@lru_cache()
+@lru_cache
 # Copied from transformers.models.gpt2.tokenization_gpt2.bytes_to_unicode
 def bytes_to_unicode():
     """

src/transformers/models/codegen/tokenization_codegen.py

@@ -42,7 +42,7 @@
 }
 
 
-@lru_cache()
+@lru_cache
 def bytes_to_unicode():
     """
     Returns list of utf-8 byte and a mapping to unicode strings. We specifically avoids mapping to whitespace/control

src/transformers/models/convnext/modeling_convnext.py

@@ -70,7 +70,7 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         return drop_path(hidden_states, self.drop_prob, self.training)
 
     def extra_repr(self) -> str:
-        return "p={}".format(self.drop_prob)
+        return f"p={self.drop_prob}"
 
 
 class ConvNextLayerNorm(nn.Module):
@@ -149,7 +149,7 @@ def __init__(self, config, dim, drop_path=0):
         self.act = ACT2FN[config.hidden_act]
         self.pwconv2 = nn.Linear(4 * dim, dim)
         self.layer_scale_parameter = (
-            nn.Parameter(config.layer_scale_init_value * torch.ones((dim)), requires_grad=True)
+            nn.Parameter(config.layer_scale_init_value * torch.ones(dim), requires_grad=True)
             if config.layer_scale_init_value > 0
             else None
         )

src/transformers/models/convnextv2/modeling_convnextv2.py

@@ -70,7 +70,7 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         return drop_path(hidden_states, self.drop_prob, self.training)
 
     def extra_repr(self) -> str:
-        return "p={}".format(self.drop_prob)
+        return f"p={self.drop_prob}"
 
 
 class ConvNextV2GRN(nn.Module):

src/transformers/models/cpm/tokenization_cpm.py

@@ -207,7 +207,7 @@ def _tokenize(self, text: str) -> list[str]:
         pieces = self.sp_model.encode(text, out_type=str)
         new_pieces = []
         for piece in pieces:
-            if len(piece) > 1 and piece[-1] == str(",") and piece[-2].isdigit():
+            if len(piece) > 1 and piece[-1] == "," and piece[-2].isdigit():
                 cur_pieces = self.sp_model.EncodeAsPieces(piece[:-1].replace(SPIECE_UNDERLINE, ""))
                 if piece[0] != SPIECE_UNDERLINE and cur_pieces[0][0] == SPIECE_UNDERLINE:
                     if len(cur_pieces[0]) == 1:

src/transformers/models/cvt/modeling_cvt.py

@@ -86,7 +86,7 @@ def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         return drop_path(hidden_states, self.drop_prob, self.training)
 
     def extra_repr(self) -> str:
-        return "p={}".format(self.drop_prob)
+        return f"p={self.drop_prob}"
 
 
 class CvtEmbeddings(nn.Module):

src/transformers/models/d_fine/modeling_d_fine.py

@@ -187,7 +187,7 @@ def forward(
             sampling_locations = reference_points[:, :, None, :, :2] + offset
         else:
             raise ValueError(
-                "Last dim of reference_points must be 2 or 4, but get {} instead.".format(reference_points.shape[-1])
+                f"Last dim of reference_points must be 2 or 4, but get {reference_points.shape[-1]} instead."
             )
 
         output = self.ms_deformable_attn_core(