Add `pyupgrade` to `pre-commit` (#8557)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

.github/prepare-nightly_version.py

@@ -11,7 +11,7 @@ now = datetime.datetime.now()
 now_date = now.strftime("%Y%m%d")
 
 print(f"prepare init '{_PATH_INFO}' - replace version by {now_date}")
-with open(_PATH_INFO, "r") as fp:
+with open(_PATH_INFO) as fp:
     init = fp.read()
 init = re.sub(r'__version__ = [\d\.\w\'"]+', f'__version__ = "{now_date}"', init)
 with open(_PATH_INFO, "w") as fp:

.github/prune-packages.py

@@ -3,7 +3,7 @@ from pprint import pprint
 
 
 def main(req_file: str, *pkgs):
-    with open(req_file, "r") as fp:
+    with open(req_file) as fp:
         lines = fp.readlines()
 
     for pkg in pkgs:

.pre-commit-config.yaml

@@ -45,6 +45,13 @@ repos:
             )$
       - id: detect-private-key
 
+  - repo: https://github.com/asottile/pyupgrade
+    rev: v2.23.0
+    hooks:
+      - id: pyupgrade
+        args: [--py36-plus]
+        name: Upgrade code
+
   - repo: https://github.com/PyCQA/isort
     rev: 5.9.2
     hooks:

docs/source/conf.py

@@ -1,4 +1,3 @@
-# -*- coding: utf-8 -*-
 #
 # Configuration file for the Sphinx documentation builder.
 #
@@ -48,7 +47,7 @@ HelperCLI.copy_notebooks(PATH_RAW_NB, PATH_IPYNB)
 
 
 def _transform_changelog(path_in: str, path_out: str) -> None:
-    with open(path_in, "r") as fp:
+    with open(path_in) as fp:
         chlog_lines = fp.readlines()
     # enrich short subsub-titles to be unique
     chlog_ver = ""
@@ -291,7 +290,7 @@ def setup(app):
 def package_list_from_file(file):
     """List up package name (not containing version and extras) from a package list file"""
     mocked_packages = []
-    with open(file, "r") as fp:
+    with open(file) as fp:
         for ln in fp.readlines():
             # Example: `tqdm>=4.41.0` => `tqdm`
             # `[` is for package with extras

pl_examples/basic_examples/dali_image_classifier.py

@@ -44,7 +44,7 @@ else:
     ops, Pipeline, DALIClassificationIterator, LastBatchPolicy = ..., ABC, ABC, ABC
 
 
-class ExternalMNISTInputIterator(object):
+class ExternalMNISTInputIterator:
     """
     This iterator class wraps torchvision's MNIST dataset and returns the images and labels in batches
     """
@@ -78,7 +78,7 @@ class ExternalSourcePipeline(Pipeline):
     """
 
     def __init__(self, batch_size, eii, num_threads, device_id):
-        super(ExternalSourcePipeline, self).__init__(batch_size, num_threads, device_id, seed=12)
+        super().__init__(batch_size, num_threads, device_id, seed=12)
         self.source = ops.ExternalSource(source=eii, num_outputs=2)
         self.build()
 

pl_examples/domain_templates/reinforce_learn_Qnet.py

@@ -66,7 +66,7 @@ class DQN(nn.Module):
             n_actions: number of discrete actions available in the environment
             hidden_size: size of hidden layers
         """
-        super(DQN, self).__init__()
+        super().__init__()
         self.net = nn.Sequential(nn.Linear(obs_size, hidden_size), nn.ReLU(), nn.Linear(hidden_size, n_actions))
 
     def forward(self, x):
@@ -106,7 +106,7 @@ class ReplayBuffer:
 
     def sample(self, batch_size: int) -> Tuple:
         indices = np.random.choice(len(self.buffer), batch_size, replace=False)
-        states, actions, rewards, dones, next_states = zip(*[self.buffer[idx] for idx in indices])
+        states, actions, rewards, dones, next_states = zip(*(self.buffer[idx] for idx in indices))
 
         return (
             np.array(states),

pytorch_lightning/callbacks/pruning.py

@@ -52,7 +52,11 @@ _PYTORCH_PRUNING_METHOD = {
 _PARAM_TUPLE = Tuple[nn.Module, str]
 _PARAM_LIST = Sequence[_PARAM_TUPLE]
 _MODULE_CONTAINERS = (LightningModule, nn.Sequential, nn.ModuleList, nn.ModuleDict)
-_LayerRef = TypedDict("_LayerRef", {"data": nn.Module, "names": List[Tuple[int, str]]})
+
+
+class _LayerRef(TypedDict):
+    data: nn.Module
+    names: List[Tuple[int, str]]
 
 
 class ModelPruning(Callback):

pytorch_lightning/core/memory.py

@@ -34,7 +34,7 @@ PARAMETER_NUM_UNITS = [" ", "K", "M", "B", "T"]
 UNKNOWN_SIZE = "?"
 
 
-class LayerSummary(object):
+class LayerSummary:
     """
     Summary class for a single layer in a :class:`~pytorch_lightning.core.lightning.LightningModule`.
     It collects the following information:
@@ -126,7 +126,7 @@ class LayerSummary(object):
         return sum(np.prod(p.shape) if not _is_lazy_weight_tensor(p) else 0 for p in self._module.parameters())
 
 
-class ModelSummary(object):
+class ModelSummary:
     """
     Generates a summary of all layers in a :class:`~pytorch_lightning.core.lightning.LightningModule`.
 

pytorch_lightning/core/mixins/hparams_mixin.py

@@ -22,7 +22,7 @@ from pytorch_lightning.utilities import AttributeDict
 from pytorch_lightning.utilities.parsing import save_hyperparameters
 
 
-class HyperparametersMixin(object):
+class HyperparametersMixin:
 
     __jit_unused_properties__ = ["hparams", "hparams_initial"]
 

pytorch_lightning/core/saving.py

@@ -44,7 +44,7 @@ if _OMEGACONF_AVAILABLE:
 CHECKPOINT_PAST_HPARAMS_KEYS = ("hparams", "module_arguments")  # used in 0.7.6
 
 
-class ModelIO(object):
+class ModelIO:
     CHECKPOINT_HYPER_PARAMS_KEY = "hyper_parameters"
     CHECKPOINT_HYPER_PARAMS_NAME = "hparams_name"
     CHECKPOINT_HYPER_PARAMS_TYPE = "hparams_type"

pytorch_lightning/loggers/base.py

@@ -249,8 +249,7 @@ class LightningLoggerBase(ABC):
                     key = str(key)
                     if isinstance(value, (MutableMapping, Namespace)):
                         value = vars(value) if isinstance(value, Namespace) else value
-                        for d in _dict_generator(value, prefixes + [key]):
-                            yield d
+                        yield from _dict_generator(value, prefixes + [key])
                     else:
                         yield prefixes + [key, value if value is not None else str(None)]
             else:
@@ -425,7 +424,7 @@ class LoggerCollection(LightningLoggerBase):
         return "_".join([str(logger.version) for logger in self._logger_iterable])
 
 
-class DummyExperiment(object):
+class DummyExperiment:
     """Dummy experiment"""
 
     def nop(self, *args, **kw):

pytorch_lightning/loggers/csv_logs.py

@@ -19,7 +19,6 @@ CSV logger for basic experiment logging that does not require opening ports
 
 """
 import csv
-import io
 import logging
 import os
 from argparse import Namespace
@@ -35,7 +34,7 @@ from pytorch_lightning.utilities.distributed import rank_zero_only
 log = logging.getLogger(__name__)
 
 
-class ExperimentWriter(object):
+class ExperimentWriter:
     r"""
     Experiment writer for CSVLogger.
 
@@ -95,7 +94,7 @@ class ExperimentWriter(object):
             last_m.update(m)
         metrics_keys = list(last_m.keys())
 
-        with io.open(self.metrics_file_path, "w", newline="") as f:
+        with open(self.metrics_file_path, "w", newline="") as f:
             self.writer = csv.DictWriter(f, fieldnames=metrics_keys)
             self.writer.writeheader()
             self.writer.writerows(self.metrics)

pytorch_lightning/loggers/wandb.py

@@ -253,7 +253,7 @@ class WandbLogger(LightningLoggerBase):
             checkpoint_callback.best_model_path: checkpoint_callback.best_model_score,
             **checkpoint_callback.best_k_models,
         }
-        checkpoints = sorted([(Path(p).stat().st_mtime, p, s) for p, s in checkpoints.items() if Path(p).is_file()])
+        checkpoints = sorted((Path(p).stat().st_mtime, p, s) for p, s in checkpoints.items() if Path(p).is_file())
         checkpoints = [
             c for c in checkpoints if c[1] not in self._logged_model_time.keys() or self._logged_model_time[c[1]] < c[0]
         ]

pytorch_lightning/overrides/torch_distributed.py

@@ -48,7 +48,7 @@ def _broadcast_object_list(object_list, src=0, group=None):
     my_rank = get_rank()
     # Serialize object_list elements to tensors on src rank.
     if my_rank == src:
-        tensor_list, size_list = zip(*[_object_to_tensor(obj) for obj in object_list])
+        tensor_list, size_list = zip(*(_object_to_tensor(obj) for obj in object_list))
         object_sizes_tensor = torch.cat(size_list)
     else:
         object_sizes_tensor = torch.LongTensor(len(object_list))

pytorch_lightning/plugins/precision/precision_plugin.py

@@ -40,8 +40,7 @@ class PrecisionPlugin(Plugin, CheckpointHooks):
         Maybe different in other precision plugins.
         """
         for group in optimizer.param_groups:
-            for p in group["params"]:
-                yield p
+            yield from group["params"]
 
     def connect(
         self, model: Module, optimizers: List[Optimizer], lr_schedulers: List[Any]

pytorch_lightning/plugins/training_type/horovod.py

@@ -200,5 +200,5 @@ class HorovodPlugin(ParallelPlugin):
 
     @staticmethod
     def _filter_named_parameters(model: nn.Module, optimizer: Optimizer) -> List[Tuple[str, nn.Parameter]]:
-        opt_params = set(p for group in optimizer.param_groups for p in group.get("params", []))
+        opt_params = {p for group in optimizer.param_groups for p in group.get("params", [])}
         return [(name, p) for name, p in model.named_parameters() if p in opt_params]

pytorch_lightning/setup_tools.py

@@ -25,7 +25,7 @@ def _load_requirements(path_dir: str, file_name: str = "requirements.txt", comme
     >>> _load_requirements(_PROJECT_ROOT)  # doctest: +ELLIPSIS +NORMALIZE_WHITESPACE
     ['numpy...', 'torch...', ...]
     """
-    with open(os.path.join(path_dir, file_name), "r") as file:
+    with open(os.path.join(path_dir, file_name)) as file:
         lines = [ln.strip() for ln in file.readlines()]
     reqs = []
     for ln in lines:

pytorch_lightning/trainer/connectors/accelerator_connector.py

@@ -80,7 +80,7 @@ if _HOROVOD_AVAILABLE:
 log = logging.getLogger(__name__)
 
 
-class AcceleratorConnector(object):
+class AcceleratorConnector:
     def __init__(
         self,
         num_processes,

pytorch_lightning/trainer/supporters.py

@@ -35,7 +35,7 @@ from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.imports import _fault_tolerant_enabled
 
 
-class TensorRunningAccum(object):
+class TensorRunningAccum:
     """Tracks a running accumulation values (min, max, mean) without graph
     references.
 
@@ -112,7 +112,7 @@ class TensorRunningAccum(object):
             return getattr(self.memory[: self.current_idx], how)()
 
 
-class PredictionCollection(object):
+class PredictionCollection:
     def __init__(self, global_rank: int, world_size: int):
         self.global_rank = global_rank
         self.world_size = world_size
@@ -170,7 +170,7 @@ class PredictionCollection(object):
                 torch.save(outputs, fp)
 
 
-class CycleIterator(object):
+class CycleIterator:
     """
     Iterator for restarting a dataloader if it runs out of samples
     """
@@ -230,7 +230,7 @@ class CycleIterator(object):
         return self.length
 
 
-class CombinedDataset(object):
+class CombinedDataset:
     """
     Combine multiple datasets and compute their statistics
     """
@@ -323,7 +323,7 @@ class DataLoaderDict(Dict):
     pass
 
 
-class CombinedLoader(object):
+class CombinedLoader:
     """
     Combines different dataloaders and allows sampling in parallel.
     Supported modes are 'min_size', which raises StopIteration after the shortest loader
@@ -510,7 +510,7 @@ class CombinedLoader(object):
         return self._calc_num_batches(self.loaders)
 
 
-class CombinedLoaderIterator(object):
+class CombinedLoaderIterator:
     """
     Custom Iterator returning data from multple loaders, and allows sampling in parallel
     """

pytorch_lightning/trainer/trainer.py

@@ -875,7 +875,7 @@ class Trainer(
         # ----------------------------
         # INSPECT THE CORE LOOPS
         # ----------------------------
-        f"""
+        fr"""
              Lightning internal flow looks like this:
         {Trainer.fit} or {Trainer.test} or {Trainer.predict}  ||
                                 |                             ||

pytorch_lightning/tuner/lr_finder.py

@@ -60,7 +60,7 @@ def _determine_lr_attr_name(trainer: "pl.Trainer", model: "pl.LightningModule")
     )
 
 
-class _LRFinder(object):
+class _LRFinder:
     """LR finder object. This object stores the results of lr_find().
 
     Args:
@@ -396,7 +396,7 @@ class _LinearLR(_LRScheduler):
     def __init__(self, optimizer: torch.optim.Optimizer, end_lr: float, num_iter: int, last_epoch: int = -1):
         self.end_lr = end_lr
         self.num_iter = num_iter
-        super(_LinearLR, self).__init__(optimizer, last_epoch)
+        super().__init__(optimizer, last_epoch)
 
     def get_lr(self):
         curr_iter = self.last_epoch + 1
@@ -435,7 +435,7 @@ class _ExponentialLR(_LRScheduler):
     def __init__(self, optimizer: torch.optim.Optimizer, end_lr: float, num_iter: int, last_epoch: int = -1):
         self.end_lr = end_lr
         self.num_iter = num_iter
-        super(_ExponentialLR, self).__init__(optimizer, last_epoch)
+        super().__init__(optimizer, last_epoch)
 
     def get_lr(self):
         curr_iter = self.last_epoch + 1

pytorch_lightning/utilities/debugging.py

@@ -34,7 +34,7 @@ def enabled_only(fn: Callable):
     return wrapped_fn
 
 
-class InternalDebugger(object):
+class InternalDebugger:
     def __init__(self, trainer):
         self.enabled = os.environ.get("PL_DEV_DEBUG", "0") == "1"
         self.trainer = trainer

pytorch_lightning/utilities/parsing.py

@@ -280,7 +280,7 @@ class AttributeDict(Dict):
     def __repr__(self) -> str:
         if not len(self):
             return ""
-        max_key_length = max([len(str(k)) for k in self])
+        max_key_length = max(len(str(k)) for k in self)
         tmp_name = "{:" + str(max_key_length + 3) + "s} {}"
         rows = [tmp_name.format(f'"{n}":', self[n]) for n in sorted(self.keys())]
         out = "\n".join(rows)

requirements/adjust_versions.py

@@ -37,7 +37,7 @@ def main(path_req: str, torch_version: Optional[str] = None) -> None:
         torch_version = torch.__version__
     assert torch_version, f"invalid torch: {torch_version}"
 
-    with open(path_req, "r") as fp:
+    with open(path_req) as fp:
         req = fp.read()
     # remove comments
     req = re.sub(rf"\s*#.*{os.linesep}", os.linesep, req)

tests/callbacks/test_early_stopping.py

@@ -275,7 +275,7 @@ def test_min_steps_override_early_stopping_functionality(tmpdir, step_freeze: in
 
     class Model(BoringModel):
         def __init__(self, step_freeze):
-            super(Model, self).__init__()
+            super().__init__()
 
             self._step_freeze = step_freeze
 

tests/callbacks/test_lr_monitor.py

@@ -351,7 +351,7 @@ def test_multiple_optimizers_basefinetuning(tmpdir):
 
     class Check(Callback):
         def on_train_epoch_start(self, trainer, pl_module) -> None:
-            num_param_groups = sum([len(opt.param_groups) for opt in trainer.optimizers])
+            num_param_groups = sum(len(opt.param_groups) for opt in trainer.optimizers)
             assert lr_monitor.lr_sch_names == ["lr-Adam", "lr-Adam-1"]
             if trainer.current_epoch == 0:
                 assert num_param_groups == 3

tests/callbacks/test_pruning.py

@@ -56,7 +56,7 @@ class TestPruningMethod(pytorch_prune.BasePruningMethod):
 
     @classmethod
     def apply(cls, module, name, amount):
-        return super(TestPruningMethod, cls).apply(module, name, amount=amount)
+        return super().apply(module, name, amount=amount)
 
 
 def train_with_pruning_callback(

tests/checkpointing/test_model_checkpoint.py

@@ -312,7 +312,7 @@ def test_model_checkpoint_to_yaml(tmpdir, save_top_k: int):
 
     path_yaml = os.path.join(tmpdir, "best_k_models.yaml")
     checkpoint.to_yaml(path_yaml)
-    d = yaml.full_load(open(path_yaml, "r"))
+    d = yaml.full_load(open(path_yaml))
     best_k = dict(checkpoint.best_k_models.items())
     assert d == best_k
 
@@ -1206,7 +1206,7 @@ def test_ckpt_version_after_rerun_same_trainer(tmpdir):
     trainer.fit(BoringModel())
 
     ckpt_range = range(mc.STARTING_VERSION, trainer.max_epochs + mc.STARTING_VERSION)
-    expected = {"test.ckpt", *[f"test-v{i}.ckpt" for i in ckpt_range]}
+    expected = {"test.ckpt", *(f"test-v{i}.ckpt" for i in ckpt_range)}
     # check best_k_models state
     assert {Path(f).name for f in mc.best_k_models} == expected
     # check created ckpts

tests/conftest.py

@@ -68,7 +68,7 @@ def pytest_pyfunc_call(pyfuncitem):
     if pyfuncitem.get_closest_marker("spawn"):
         testfunction = pyfuncitem.obj
         funcargs = pyfuncitem.funcargs
-        testargs = tuple([funcargs[arg] for arg in pyfuncitem._fixtureinfo.argnames])
+        testargs = tuple(funcargs[arg] for arg in pyfuncitem._fixtureinfo.argnames)
 
         mp.spawn(wraps, (testfunction, testargs))
         return True

tests/core/test_lightning_module.py

@@ -263,10 +263,8 @@ def test_toggle_untoggle_3_optimizers_shared_parameters(tmpdir):
 
         @staticmethod
         def combine_generators(gen_1, gen_2):
-            for p in gen_1:
-                yield p
-            for p in gen_2:
-                yield p
+            yield from gen_1
+            yield from gen_2
 
         def configure_optimizers(self):
             optimizer_1 = SGD(self.combine_generators(self.layer_1.parameters(), self.layer_2.parameters()), lr=0.1)

tests/core/test_lightning_optimizer.py

@@ -302,7 +302,7 @@ class OptimizerWithHooks(Optimizer):
             d = {"params": params, "mod": mod, "layer_type": mod_class}
             self.params.append(d)
 
-        super(OptimizerWithHooks, self).__init__(self.params, {"lr": 0.01})
+        super().__init__(self.params, {"lr": 0.01})
 
     def _save_input(self, mod, i):
         """Saves input of layer"""

tests/loggers/test_csv.py

@@ -79,7 +79,7 @@ def test_file_logger_log_metrics(tmpdir, step_idx):
     logger.save()
 
     path_csv = os.path.join(logger.log_dir, ExperimentWriter.NAME_METRICS_FILE)
-    with open(path_csv, "r") as fp:
+    with open(path_csv) as fp:
         lines = fp.readlines()
     assert len(lines) == 2
     assert all(n in lines[0] for n in metrics)

tests/models/test_gpu.py

@@ -365,7 +365,7 @@ def test_non_blocking():
         batch = trainer.accelerator.batch_to_device(batch, torch.device("cuda:0"))
         mocked.assert_called_with(torch.device("cuda", 0), non_blocking=True)
 
-    class BatchObject(object):
+    class BatchObject:
         def to(self, *args, **kwargs):
             pass
 

tests/models/test_horovod.py

@@ -221,11 +221,11 @@ def test_horovod_transfer_batch_to_gpu(tmpdir):
     class TestTrainingStepModel(BoringModel):
         def training_step(self, batch, *args, **kwargs):
             assert str(batch.device) != "cpu"
-            return super(TestTrainingStepModel, self).training_step(batch, *args, **kwargs)
+            return super().training_step(batch, *args, **kwargs)
 
         def validation_step(self, batch, *args, **kwargs):
             assert str(batch.device) != "cpu"
-            return super(TestTrainingStepModel, self).validation_step(batch, *args, **kwargs)
+            return super().validation_step(batch, *args, **kwargs)
 
     model = TestTrainingStepModel()
 
@@ -269,7 +269,7 @@ def test_horovod_multi_optimizer(tmpdir):
         return set(list(model.parameters()))
 
     def get_optimizer_params(optimizer):
-        return set([p for group in optimizer.param_groups for p in group.get("params", [])])
+        return {p for group in optimizer.param_groups for p in group.get("params", [])}
 
     assert get_model_params(model.generator) != get_model_params(model.discriminator)
     assert get_model_params(model.generator) == get_optimizer_params(trainer.optimizers[0])

tests/profiler/test_profiler.py

@@ -36,7 +36,7 @@ PROFILER_OVERHEAD_MAX_TOLERANCE = 0.0005
 
 
 def _get_python_cprofile_total_duration(profile):
-    return sum([x.inlinetime for x in profile.getstats()])
+    return sum(x.inlinetime for x in profile.getstats())
 
 
 def _sleep_generator(durations):
@@ -322,7 +322,7 @@ def test_pytorch_profiler_trainer_test(tmpdir):
     assert path.read_text("utf-8")
 
     if _KINETO_AVAILABLE:
-        files = sorted([file for file in os.listdir(tmpdir) if file.endswith(".json")])
+        files = sorted(file for file in os.listdir(tmpdir) if file.endswith(".json"))
         assert any(f"test-{pytorch_profiler.filename}" in f for f in files)
         path = pytorch_profiler.dirpath / f"test-{pytorch_profiler.filename}.txt"
         assert path.read_text("utf-8")

tests/trainer/logging_/test_logger_connector.py

@@ -29,7 +29,7 @@ from tests.helpers.runif import RunIf
 
 
 def test_fx_validator(tmpdir):
-    funcs_name = sorted([f for f in dir(Callback) if not f.startswith("_")])
+    funcs_name = sorted(f for f in dir(Callback) if not f.startswith("_"))
 
     callbacks_func = [
         "on_before_backward",

tests/trainer/test_supporters.py

@@ -169,7 +169,7 @@ def test_combined_loader_dict_min_size():
 
     combined_loader = CombinedLoader(loaders, "min_size")
 
-    assert len(combined_loader) == min([len(v) for v in loaders.values()])
+    assert len(combined_loader) == min(len(v) for v in loaders.values())
 
     for idx, item in enumerate(combined_loader):
         assert isinstance(item, dict)
@@ -188,7 +188,7 @@ def test_combined_loader_dict_max_size_cycle():
 
     combined_loader = CombinedLoader(loaders, "max_size_cycle")
 
-    assert len(combined_loader) == max([len(v) for v in loaders.values()])
+    assert len(combined_loader) == max(len(v) for v in loaders.values())
 
     for idx, item in enumerate(combined_loader):
         assert isinstance(item, dict)
@@ -207,7 +207,7 @@ def test_combined_loader_sequence_min_size():
 
     combined_loader = CombinedLoader(loaders, "min_size")
 
-    assert len(combined_loader) == min([len(v) for v in loaders])
+    assert len(combined_loader) == min(len(v) for v in loaders)
 
     for idx, item in enumerate(combined_loader):
         assert isinstance(item, Sequence)
@@ -225,7 +225,7 @@ def test_combined_loader_sequence_max_size_cycle():
 
     combined_loader = CombinedLoader(loaders, "max_size_cycle")
 
-    assert len(combined_loader) == max([len(v) for v in loaders])
+    assert len(combined_loader) == max(len(v) for v in loaders)
 
     for idx, item in enumerate(combined_loader):
         assert isinstance(item, Sequence)

tests/trainer/test_trainer.py

@@ -879,7 +879,7 @@ def test_gradient_clipping(tmpdir):
         ret_val = old_training_step_and_backward(split_batch, batch_idx, opt_idx, optimizer, hiddens)
         parameters = model.parameters()
         grad_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), 2) for p in parameters]), 2)
-        assert (grad_norm - 1.0).abs() < 0.01, "Gradient norm != 1.0: {grad_norm}".format(grad_norm=grad_norm)
+        assert (grad_norm - 1.0).abs() < 0.01, f"Gradient norm != 1.0: {grad_norm}"
 
         return ret_val
 
@@ -952,7 +952,7 @@ def test_gradient_clipping_fp16(tmpdir):
         ret_val = old_training_step_and_backward(split_batch, batch_idx, opt_idx, optimizer, hiddens)
         parameters = model.parameters()
         grad_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), 2) for p in parameters]), 2)
-        assert (grad_norm - 1.0).abs() < 0.01, "Gradient norm != 1.0: {grad_norm}".format(grad_norm=grad_norm)
+        assert (grad_norm - 1.0).abs() < 0.01, f"Gradient norm != 1.0: {grad_norm}"
 
         return ret_val