use isinstance instead of type comparison

examples/flax/image-captioning/run_image_captioning_flax.py

@@ -1174,7 +1174,7 @@ def evaluation_loop(
                 with open(output_file, "w", encoding="UTF-8") as fp:
                     json.dump(generations, fp, ensure_ascii=False, indent=4)
 
-    def evaluate(rng: jax.random.PRNGKey, dataset: Dataset, ckpt_dir: str = ""):
+    def evaluation(rng: jax.random.PRNGKey, dataset: Dataset, ckpt_dir: str = ""):
         evaluation_loop(rng, dataset, metric_key_prefix="eval", ckpt_dir=ckpt_dir)
 
     def predict(rng: jax.random.PRNGKey, dataset: Dataset):
@@ -1247,7 +1247,7 @@ def predict(rng: jax.random.PRNGKey, dataset: Dataset):
                 ):
                     ckpt_dir = f"ckpt_epoch_{epoch + 1}_step_{cur_step}"
                     commit_msg = f"Saving weights and logs of epoch {epoch + 1} - step {cur_step}"
-                    evaluate(input_rng, eval_dataset, ckpt_dir)
+                    evaluation(input_rng, eval_dataset, ckpt_dir)
                     save_ckpt(ckpt_dir=ckpt_dir, commit_msg=commit_msg)
 
             # ======================== Epoch End ==============================
@@ -1270,7 +1270,7 @@ def predict(rng: jax.random.PRNGKey, dataset: Dataset):
             if training_args.do_eval and (training_args.eval_steps is None or training_args.eval_steps <= 0):
                 ckpt_dir = f"ckpt_epoch_{epoch + 1}_step_{cur_step}"
                 commit_msg = f"Saving weights and logs of epoch {epoch + 1} - step {cur_step}"
-                evaluate(input_rng, eval_dataset, ckpt_dir)
+                evaluation(input_rng, eval_dataset, ckpt_dir)
                 save_ckpt(ckpt_dir=ckpt_dir, commit_msg=commit_msg)
 
     # ======================== Evaluating | Predicting ==============================

examples/flax/text-classification/run_flax_glue.py

@@ -212,7 +212,7 @@ def __post_init__(self):
             if self.validation_file is not None:
                 extension = self.validation_file.split(".")[-1]
                 assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
-        self.task_name = self.task_name.lower() if type(self.task_name) == str else self.task_name
+        self.task_name = self.task_name.lower() if isinstance(self.task_name, str) else self.task_name
 
 
 def create_train_state(

examples/legacy/pytorch-lightning/run_glue.py

@@ -23,7 +23,7 @@ class GLUETransformer(BaseTransformer):
     mode = "sequence-classification"
 
     def __init__(self, hparams):
-        if type(hparams) == dict:
+        if isinstance(hparams, dict):
             hparams = Namespace(**hparams)
         hparams.glue_output_mode = glue_output_modes[hparams.task]
         num_labels = glue_tasks_num_labels[hparams.task]

examples/legacy/pytorch-lightning/run_ner.py

@@ -25,7 +25,7 @@ class NERTransformer(BaseTransformer):
     mode = "token-classification"
 
     def __init__(self, hparams):
-        if type(hparams) == dict:
+        if isinstance(hparams, dict):
             hparams = Namespace(**hparams)
         module = import_module("tasks")
         try:

examples/research_projects/deebert/src/modeling_highway_bert.py

@@ -32,7 +32,7 @@ def __init__(self, config):
         self.early_exit_entropy = [-1 for _ in range(config.num_hidden_layers)]
 
     def set_early_exit_entropy(self, x):
-        if (type(x) is float) or (type(x) is int):
+        if isinstance(x, (float, int)):
             for i in range(len(self.early_exit_entropy)):
                 self.early_exit_entropy[i] = x
         else:

examples/research_projects/quantization-qdqbert/quant_trainer.py

@@ -239,7 +239,7 @@ def print_model_summary(model, name_width=25, line_width=180, ignore=None):
             continue
         if type(mod) in ignore:
             continue
-        if [True for s in ignore if type(s) is str and s in name]:
+        if [True for s in ignore if isinstance(s, str) and s in name]:
             continue
         act_str = f"Act:{input_q.extra_repr()}"
         wgt_str = f"Wgt:{weight_q.extra_repr()}"

setup.py

@@ -156,7 +156,7 @@
     "rhoknp>=1.1.0,<1.3.1",
     "rjieba",
     "rouge-score!=0.0.7,!=0.0.8,!=0.1,!=0.1.1",
-    "ruff>=0.0.241,<=0.0.259",
+    "ruff>=0.0.241",
     "sacrebleu>=1.4.12,<2.0.0",
     "sacremoses",
     "safetensors>=0.3.1",

src/transformers/data/data_collator.py

@@ -121,7 +121,7 @@ def torch_default_data_collator(features: List[InputDataClass]) -> Dict[str, Any
         if isinstance(first["label_ids"], torch.Tensor):
             batch["labels"] = torch.stack([f["label_ids"] for f in features])
         else:
-            dtype = torch.long if type(first["label_ids"][0]) is int else torch.float
+            dtype = torch.long if isinstance(first["label_ids"][0], int) else torch.float
             batch["labels"] = torch.tensor([f["label_ids"] for f in features], dtype=dtype)
 
     # Handling of all other possible keys.
@@ -196,7 +196,7 @@ def numpy_default_data_collator(features: List[InputDataClass]) -> Dict[str, Any
         if isinstance(first["label_ids"], np.ndarray):
             batch["labels"] = np.stack([f["label_ids"] for f in features])
         else:
-            dtype = np.int64 if type(first["label_ids"][0]) is int else np.float32
+            dtype = np.int64 if isinstance(first["label_ids"][0], int) else np.float32
             batch["labels"] = np.array([f["label_ids"] for f in features], dtype=dtype)
 
     # Handling of all other possible keys.

src/transformers/dependency_versions_table.py

@@ -62,7 +62,7 @@
     "rhoknp": "rhoknp>=1.1.0,<1.3.1",
     "rjieba": "rjieba",
     "rouge-score": "rouge-score!=0.0.7,!=0.0.8,!=0.1,!=0.1.1",
-    "ruff": "ruff>=0.0.241,<=0.0.259",
+    "ruff": "ruff>=0.0.241",
     "sacrebleu": "sacrebleu>=1.4.12,<2.0.0",
     "sacremoses": "sacremoses",
     "safetensors": "safetensors>=0.3.1",

src/transformers/models/blip/modeling_blip_text.py

@@ -747,13 +747,13 @@ def forward(
         # If a 2D or 3D attention mask is provided for the cross-attention
         # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
         if encoder_hidden_states is not None:
-            if type(encoder_hidden_states) == list:
+            if isinstance(encoder_hidden_states, list):
                 encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states[0].size()
             else:
                 encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
             encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
 
-            if type(encoder_attention_mask) == list:
+            if isinstance(encoder_attention_mask, list):
                 encoder_extended_attention_mask = [self.invert_attention_mask(mask) for mask in encoder_attention_mask]
             elif encoder_attention_mask is None:
                 encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)

src/transformers/models/blip/modeling_tf_blip_text.py

@@ -741,13 +741,13 @@ def call(
         # If a 2D or 3D attention mask is provided for the cross-attention
         # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
         if encoder_hidden_states is not None:
-            if type(encoder_hidden_states) == list:
+            if isinstance(encoder_hidden_states, list):
                 encoder_batch_size, encoder_sequence_length, _ = shape_list(encoder_hidden_states[0])
             else:
                 encoder_batch_size, encoder_sequence_length, _ = shape_list(encoder_hidden_states)
             encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
 
-            if type(encoder_attention_mask) == list:
+            if isinstance(encoder_attention_mask, list):
                 encoder_extended_attention_mask = [invert_attention_mask(mask) for mask in encoder_attention_mask]
             elif encoder_attention_mask is None:
                 encoder_attention_mask = tf.ones(encoder_hidden_shape)

src/transformers/models/blip_2/modeling_blip_2.py

@@ -1140,13 +1140,13 @@ def forward(
         # If a 2D or 3D attention mask is provided for the cross-attention
         # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
         if encoder_hidden_states is not None:
-            if type(encoder_hidden_states) == list:
+            if isinstance(encoder_hidden_states, list):
                 encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states[0].size()
             else:
                 encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
             encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
 
-            if type(encoder_attention_mask) == list:
+            if isinstance(encoder_attention_mask, list):
                 encoder_extended_attention_mask = [self.invert_attention_mask(mask) for mask in encoder_attention_mask]
             elif encoder_attention_mask is None:
                 encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)

src/transformers/models/deberta/configuration_deberta.py

@@ -148,7 +148,7 @@ def __init__(
         self.position_biased_input = position_biased_input
 
         # Backwards compatibility
-        if type(pos_att_type) == str:
+        if isinstance(pos_att_type, str):
             pos_att_type = [x.strip() for x in pos_att_type.lower().split("|")]
 
         self.pos_att_type = pos_att_type

src/transformers/models/deberta_v2/configuration_deberta_v2.py

@@ -150,7 +150,7 @@ def __init__(
         self.position_biased_input = position_biased_input
 
         # Backwards compatibility
-        if type(pos_att_type) == str:
+        if isinstance(pos_att_type, str):
             pos_att_type = [x.strip() for x in pos_att_type.lower().split("|")]
 
         self.pos_att_type = pos_att_type

src/transformers/models/esm/modeling_esmfold.py

@@ -229,7 +229,7 @@ def dict_multimap(fn, dicts):
     new_dict = {}
     for k, v in first.items():
         all_v = [d[k] for d in dicts]
-        if type(v) is dict:
+        if isinstance(v, dict):
             new_dict[k] = dict_multimap(fn, all_v)
         else:
             new_dict[k] = fn(all_v)
@@ -1060,7 +1060,7 @@ def __init__(self, r: float, batch_dim: Union[int, List[int]]):
         super().__init__()
 
         self.r = r
-        if type(batch_dim) == int:
+        if isinstance(batch_dim, int):
             batch_dim = [batch_dim]
         self.batch_dim = batch_dim
         self.dropout = nn.Dropout(self.r)
@@ -2254,7 +2254,7 @@ def infer(
         seqs: Union[str, List[str]],
         position_ids=None,
     ):
-        if type(seqs) is str:
+        if isinstance(seqs, str):
             lst = [seqs]
         else:
             lst = seqs
@@ -2312,7 +2312,7 @@ def output_to_pdb(output: Dict) -> List[str]:
 
     def infer_pdb(self, seqs, *args, **kwargs) -> str:
         """Returns the pdb (file) string from the model given an input sequence."""
-        assert type(seqs) is str
+        assert isinstance(seqs, str)
         output = self.infer(seqs, *args, **kwargs)
         return self.output_to_pdb(output)[0]
 

src/transformers/models/gptsan_japanese/modeling_gptsan_japanese.py

@@ -1288,7 +1288,7 @@ def prepare_inputs_for_generation(
         past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None,
         **kwargs,
     ):
-        if type(spout) is list:
+        if isinstance(spout, list):
             spout = torch.tensor(spout).float()
             if input_ids is not None:
                 spout = spout.to(input_ids.device)

src/transformers/models/gptsan_japanese/tokenization_gptsan_japanese.py

@@ -375,7 +375,7 @@ def _batch_encode_plus(
         verbose: bool = True,
     ) -> BatchEncoding:
         # This tokenizer converts input text pairs into Prefix input and subsequent input
-        if type(batch_text_or_text_pairs[0]) is tuple or type(batch_text_or_text_pairs[0]) is list:
+        if type(batch_text_or_text_pairs[0]) is tuple or isinstance(batch_text_or_text_pairs[0], list):
             # As a single text with an explicit un-prefix position
             batch_prefix_texts = []
             for pref, txt in batch_text_or_text_pairs:

src/transformers/models/instructblip/modeling_instructblip.py

@@ -1193,13 +1193,13 @@ def forward(
         # If a 2D or 3D attention mask is provided for the cross-attention
         # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
         if encoder_hidden_states is not None:
-            if type(encoder_hidden_states) == list:
+            if isinstance(encoder_hidden_states, list):
                 encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states[0].size()
             else:
                 encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
             encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
 
-            if type(encoder_attention_mask) == list:
+            if isinstance(encoder_attention_mask, list):
                 encoder_extended_attention_mask = [self.invert_attention_mask(mask) for mask in encoder_attention_mask]
             elif encoder_attention_mask is None:
                 encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)

src/transformers/models/nougat/tokenization_nougat_fast.py

@@ -251,7 +251,7 @@ def remove_numbers(lines):
     def _clean(s):
         return re.sub(r"(?:[\d_]|\*\*)", "", s).strip()
 
-    if type(lines) is str:
+    if isinstance(lines, str):
         return _clean(lines)
     out = []
     for l in lines:

src/transformers/models/regnet/convert_regnet_to_pytorch.py

@@ -192,7 +192,7 @@ def convert_weight_and_push(
     )
 
     from_output = from_model(x)
-    from_output = from_output[-1] if type(from_output) is list else from_output
+    from_output = from_output[-1] if isinstance(from_output, list) else from_output
 
     # now since I don't want to use any config files, vissl seer model doesn't actually have an head, so let's just check the last hidden state
     if "seer" in name and "in1k" in name:

src/transformers/models/transfo_xl/tokenization_transfo_xl.py

@@ -212,7 +212,7 @@ def __init__(
                 # Loading a torch-saved transfo-xl vocab dict with pickle results in an integer
                 # Entering this if statement means that we tried to load a torch-saved file with pickle, and we failed.
                 # We therefore load it with torch, if it's available.
-                if type(vocab_dict) == int:
+                if isinstance(vocab_dict, int):
                     if not is_torch_available():
                         raise ImportError(
                             "Not trying to load dict with PyTorch as you need to install pytorch to load "

src/transformers/models/yoso/modeling_yoso.py

@@ -88,7 +88,7 @@ def to_contiguous(input_tensors):
 
 
 def normalize(input_tensors):
-    if type(input_tensors) is list:
+    if isinstance(input_tensors, list):
         out = []
         for tensor in input_tensors:
             out.append(nn.functional.normalize(tensor, p=2, dim=-1))

src/transformers/trainer_pt_utils.py

@@ -896,7 +896,7 @@ def metrics_format(self, metrics: Dict[str, float]) -> Dict[str, float]:
             metrics_copy[k] = _secs2timedelta(v)
         elif k == "total_flos":
             metrics_copy[k] = f"{ int(v) >> 30 }GF"
-        elif type(metrics_copy[k]) == float:
+        elif isinstance(metrics_copy[k], float):
             metrics_copy[k] = round(v, 4)
 
     return metrics_copy

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/configuration_{{cookiecutter.lowercase_modelname}}.py

@@ -137,7 +137,7 @@ class {{cookiecutter.camelcase_modelname}}Config(PretrainedConfig):
     {% else -%}
     keys_to_ignore_at_inference = ["past_key_values"]
     {% endif -%}
-    
+
     {% if cookiecutter.is_encoder_decoder_model == "False" %}
     {%- else %}
     attribute_map = {
@@ -238,4 +238,3 @@ def __init__(
             **kwargs
         )
 
-

templates/adding_a_new_model/cookiecutter-template-{{cookiecutter.modelname}}/test_modeling_flax_{{cookiecutter.lowercase_modelname}}.py

@@ -541,7 +541,7 @@ def prepare_{{cookiecutter.lowercase_modelname}}_inputs_dict(
 class Flax{{cookiecutter.camelcase_modelname}}ModelTest(FlaxModelTesterMixin, unittest.TestCase):
     all_model_classes = (
         (
-            Flax{{cookiecutter.camelcase_modelname}}ForConditionalGeneration, 
+            Flax{{cookiecutter.camelcase_modelname}}ForConditionalGeneration,
             Flax{{cookiecutter.camelcase_modelname}}ForQuestionAnswering,
             Flax{{cookiecutter.camelcase_modelname}}ForSequenceClassification,
             Flax{{cookiecutter.camelcase_modelname}}Model,

tests/models/encoder_decoder/test_modeling_tf_encoder_decoder.py

@@ -474,7 +474,7 @@ def check_pt_tf_outputs(self, tf_outputs, pt_outputs, model_class, tol=1e-5, nam
     def prepare_pt_inputs_from_tf_inputs(self, tf_inputs_dict):
         pt_inputs_dict = {}
         for name, key in tf_inputs_dict.items():
-            if type(key) == bool:
+            if isinstance(key, bool):
                 pt_inputs_dict[name] = key
             elif name == "input_values":
                 pt_inputs_dict[name] = torch.from_numpy(key.numpy()).to(torch.float32)

tests/models/lxmert/test_modeling_lxmert.py

@@ -751,7 +751,7 @@ def prepare_tf_inputs_from_pt_inputs(self, pt_inputs_dict):
                 tf_inputs_dict[key] = self.prepare_pt_inputs_from_tf_inputs(value)
             elif isinstance(value, (list, tuple)):
                 tf_inputs_dict[key] = (self.prepare_pt_inputs_from_tf_inputs(iter_value) for iter_value in value)
-            elif type(value) == bool:
+            elif isinstance(value, bool):
                 tf_inputs_dict[key] = value
             elif key == "input_values":
                 tf_inputs_dict[key] = tf.convert_to_tensor(value.cpu().numpy(), dtype=tf.float32)

tests/models/lxmert/test_modeling_tf_lxmert.py

@@ -499,7 +499,7 @@ def prepare_pt_inputs_from_tf_inputs(self, tf_inputs_dict):
                 pt_inputs_dict[key] = self.prepare_pt_inputs_from_tf_inputs(value)
             elif isinstance(value, (list, tuple)):
                 pt_inputs_dict[key] = (self.prepare_pt_inputs_from_tf_inputs(iter_value) for iter_value in value)
-            elif type(key) == bool:
+            elif isinstance(key, bool):
                 pt_inputs_dict[key] = value
             elif key == "input_values":
                 pt_inputs_dict[key] = torch.from_numpy(value.numpy()).to(torch.float32)

tests/models/vision_encoder_decoder/test_modeling_tf_vision_encoder_decoder.py

@@ -407,7 +407,7 @@ def check_pt_tf_outputs(self, tf_outputs, pt_outputs, model_class, tol=1e-5, nam
     def prepare_pt_inputs_from_tf_inputs(self, tf_inputs_dict):
         pt_inputs_dict = {}
         for name, key in tf_inputs_dict.items():
-            if type(key) == bool:
+            if isinstance(key, bool):
                 pt_inputs_dict[name] = key
             elif name == "input_values":
                 pt_inputs_dict[name] = torch.from_numpy(key.numpy()).to(torch.float32)

tests/test_modeling_common.py

@@ -1990,7 +1990,7 @@ def prepare_tf_inputs_from_pt_inputs(self, pt_inputs_dict):
         tf_inputs_dict = {}
         for key, tensor in pt_inputs_dict.items():
             # skip key that does not exist in tf
-            if type(tensor) == bool:
+            if isinstance(tensor, bool):
                 tf_inputs_dict[key] = tensor
             elif key == "input_values":
                 tf_inputs_dict[key] = tf.convert_to_tensor(tensor.cpu().numpy(), dtype=tf.float32)

tests/test_modeling_tf_common.py

@@ -576,7 +576,7 @@ def check_pt_tf_outputs(self, tf_outputs, pt_outputs, model_class, tol=1e-5, nam
     def prepare_pt_inputs_from_tf_inputs(self, tf_inputs_dict):
         pt_inputs_dict = {}
         for name, key in tf_inputs_dict.items():
-            if type(key) == bool:
+            if isinstance(key, bool):
                 pt_inputs_dict[name] = key
             elif name == "input_values":
                 pt_inputs_dict[name] = torch.from_numpy(key.numpy()).to(torch.float32)

utils/notification_service.py

@@ -20,7 +20,7 @@
 import os
 import re
 import sys
-import time
+from time import sleep
 from typing import Dict, List, Optional, Union
 
 import requests
@@ -649,7 +649,7 @@ def post_reply(self):
                         thread_ts=self.thread_ts["ts"],
                     )
 
-                    time.sleep(1)
+                    sleep(1)
 
         for job, job_result in self.additional_results.items():
             if len(job_result["failures"]):
@@ -672,7 +672,7 @@ def post_reply(self):
                         thread_ts=self.thread_ts["ts"],
                     )
 
-                    time.sleep(1)
+                    sleep(1)
 
 
 def retrieve_artifact(artifact_path: str, gpu: Optional[str]):