Lightning-Universe · Borda · Mar 28, 2023 · Jan 3, 2023 · Jan 3, 2023 · Mar 18, 2023
@@ -29,7 +29,7 @@ repos:
  name: Upgrade code
 
  - repo: https://github.com/PyCQA/docformatter
- rev: v1.5.0
+ rev: v1.5.1
  hooks:
  - id: docformatter
  args: [--in-place, --wrap-summaries=115, --wrap-descriptions=120]
@@ -45,12 +45,12 @@ repos:
  exclude: CHANGELOG.md
 
  - repo: https://github.com/PyCQA/isort
- rev: 5.11.2
+ rev: 5.12.0
  hooks:
  - id: isort
 
  - repo: https://github.com/psf/black
- rev: 22.12.0
+ rev: 23.1.0
  hooks:
  - id: black
  name: Format code

@@ -22,7 +22,6 @@
 
 @under_review()
 class DataMonitorBase(Callback):
-
  supported_loggers = (
  TensorBoardLogger,
  WandbLogger,
@@ -113,7 +112,6 @@ def _is_logger_available(self, logger: LightningLoggerBase) -> bool:
 
 @under_review()
 class ModuleDataMonitor(DataMonitorBase):
-
  GROUP_NAME_INPUT = "input"
  GROUP_NAME_OUTPUT = "output"
 
@@ -199,7 +197,6 @@ def hook(_: Module, inp: Sequence, out: Sequence) -> None:
 
 @under_review()
 class TrainingDataMonitor(DataMonitorBase):
-
  GROUP_NAME = "training_step"
 
  def __init__(self, log_every_n_steps: int = None):

@@ -103,7 +103,6 @@ def runner(self, device: torch.device) -> Tuple[Experience]:
 
  # step through each env
  for env_idx, (env, action) in enumerate(zip(self.pool, actions)):
-
  exp = self.env_step(env_idx, env, action)
  history = self.histories[env_idx]
  history.append(exp)

@@ -19,7 +19,6 @@
 
 @under_review()
 class KittiDataModule(LightningDataModule):
-
  name = "kitti"
 
  def __init__(

@@ -117,7 +117,6 @@ def __init__(
  *args,
  **kwargs,
  ) -> None:
-
  super().__init__(*args, **kwargs)
  self.num_workers = num_workers
  self.batch_size = batch_size

@@ -18,7 +18,6 @@
 
 @under_review()
 class SSLImagenetDataModule(LightningDataModule): # pragma: no cover
-
  name = "imagenet"
 
  def __init__(

@@ -8,7 +8,6 @@
 
 
 class VisionDataModule(LightningDataModule):
-
  EXTRA_ARGS: dict = {}
  name: str = ""
  #: Dataset class to use

@@ -16,7 +16,6 @@
 
 @under_review()
 class LightDataset(ABC, Dataset):
-
  data: Tensor
  targets: Tensor
  normalize: tuple

@@ -30,7 +30,6 @@ def generate_train_val_split(cls, examples, labels, pct_val):
 
  cts = {x: 0 for x in range(nb_classes)}
  for img, class_idx in zip(examples, labels):
-
  # allow labeled
  if cts[class_idx] < nb_val_images:
  val_x.append(img)
@@ -60,7 +59,6 @@ def select_nb_imgs_per_class(cls, examples, labels, nb_imgs_in_val):
 
  cts = {x: 0 for x in range(nb_classes)}
  for img_name, class_idx in zip(examples, labels):
-
  # allow labeled
  if cts[class_idx] < nb_imgs_in_val:
  labeled.append(img_name)
@@ -76,7 +74,6 @@ def select_nb_imgs_per_class(cls, examples, labels, nb_imgs_in_val):
 
  @classmethod
  def deterministic_shuffle(cls, x, y):
-
  n = len(x)
  idxs = list(range(0, n))
  np.random.seed(1234)

@@ -140,11 +140,8 @@ def forward(self, anchor_representations, positive_representations, mask_mat):
 
  # trick 2: tanh clip
  raw_scores = tanh_clip(raw_scores, clip_val=self.tclip)
- """
- pos_scores includes scores for all the positive samples
- neg_scores includes scores for all the negative samples, with
- scores for positive samples set to the min score (-self.tclip here)
- """
+ """pos_scores includes scores for all the positive samples neg_scores includes scores for all the negative
+ samples, with scores for positive samples set to the min score (-self.tclip here)"""
  # ----------------------
  # EXTRACT POSITIVE SCORES
  # use the index mask to pull all the diagonals which are b1 x b1
@@ -337,8 +334,7 @@ def forward(self, anchor_maps, positive_maps):
 
  regularizer = 0
  losses = []
- for (ai, pi) in self.map_indexes:
-
+ for ai, pi in self.map_indexes:
  # choose a random map
  if ai == -1:
  ai = np.random.randint(0, len(anchor_maps))

@@ -113,7 +113,6 @@ def forward(self, x):
  return self.model(x)
 
  def training_step(self, batch, batch_idx):
-
  images, targets = batch
  targets = [{k: v for k, v in t.items()} for t in targets]
 

@@ -9,7 +9,6 @@
 from pl_bolts.utils.warnings import warn_missing_pkg
 
 if _TORCHVISION_AVAILABLE:
-
  from torchvision.models.detection.retinanet import RetinaNet as torchvision_RetinaNet
  from torchvision.models.detection.retinanet import RetinaNetHead, retinanet_resnet50_fpn
  from torchvision.ops import box_iou
@@ -97,7 +96,6 @@ def forward(self, x):
  return self.model(x)
 
  def training_step(self, batch, batch_idx):
-
  images, targets = batch
  targets = [{k: v for k, v in t.items()} for t in targets]
 

@@ -18,7 +18,6 @@ def _weights_init(m):
 @under_review()
 class Pix2Pix(LightningModule):
  def __init__(self, in_channels, out_channels, learning_rate=0.0002, lambda_recon=200):
-
  super().__init__()
  self.save_hyperparameters()
 

@@ -27,7 +27,6 @@ class LitMNIST(LightningModule):
  """
 
  def __init__(self, hidden_dim: int = 128, learning_rate: float = 1e-3, **kwargs: Any) -> None:
-
  if not _TORCHVISION_AVAILABLE: # pragma: no cover
  raise ModuleNotFoundError("You want to use `torchvision` which is not installed yet.")
 

@@ -91,7 +91,13 @@ def train_batch(
  states, actions, rewards, dones, new_states = samples
 
  for idx, _ in enumerate(dones):
- yield (states[idx], actions[idx], rewards[idx], dones[idx], new_states[idx],), indices[
+ yield (
+ states[idx],
+ actions[idx],
+ rewards[idx],
+ dones[idx],
+ new_states[idx],
+ ), indices[
  idx
  ], weights[idx]
 

@@ -173,7 +173,6 @@ def train_batch(
  """
 
  while True:
-
  action = self.agent(self.state, self.device)
 
  next_state, reward, done, _ = self.env.step(action[0])

@@ -132,7 +132,6 @@ def train_batch(
  """
 
  while True:
-
  action = self.agent(self.state, self.device)
 
  next_state, reward, done, _ = self.env.step(action[0])

@@ -74,7 +74,6 @@ def __init__(
  initial_tau: float = 0.996,
  **kwargs: Any,
  ) -> None:
-
  super().__init__()
  self.save_hyperparameters(ignore="base_encoder")
 

@@ -18,7 +18,6 @@ class MLP(nn.Module):
  """
 
  def __init__(self, input_dim: int = 2048, hidden_dim: int = 4096, output_dim: int = 256) -> None:
-
  super().__init__()
 
  self.model = nn.Sequential(
@@ -53,7 +52,6 @@ def __init__(
  projector_hidden_dim: int = 4096,
  projector_out_dim: int = 256,
  ) -> None:
-
  super().__init__()
 
  if isinstance(encoder, str):

@@ -225,7 +225,6 @@ def forward(self, img_q, img_k, queue):
 
  # compute key features
  with torch.no_grad(): # no gradient to keys
-
  # shuffle for making use of BN
  if self._use_ddp(self.trainer):
  img_k, idx_unshuffle = self._batch_shuffle_ddp(img_k)

@@ -31,7 +31,6 @@ class SimCLRTrainDataTransform:
  def __init__(
  self, input_height: int = 224, gaussian_blur: bool = True, jitter_strength: float = 1.0, normalize=None
  ) -> None:
-
  if not _TORCHVISION_AVAILABLE: # pragma: no cover
  raise ModuleNotFoundError("You want to use `transforms` from `torchvision` which is not installed yet.")
 
@@ -140,7 +139,6 @@ class SimCLRFinetuneTransform(SimCLRTrainDataTransform):
  def __init__(
  self, input_height: int = 224, jitter_strength: float = 1.0, normalize=None, eval_transform: bool = False
  ) -> None:
-
  super().__init__(
  normalize=normalize, input_height=input_height, gaussian_blur=None, jitter_strength=jitter_strength
  )

@@ -75,7 +75,6 @@ def __init__(
  exclude_bn_bias: bool = False,
  **kwargs,
  ) -> None:
-
  super().__init__()
  self.save_hyperparameters(ignore="base_encoder")
 

@@ -130,7 +130,6 @@ class SwAVFinetuneTransform:
  def __init__(
  self, input_height: int = 224, jitter_strength: float = 1.0, normalize=None, eval_transform: bool = False
  ) -> None:
-
  self.jitter_strength = jitter_strength
  self.input_height = input_height
  self.normalize = normalize

@@ -32,7 +32,6 @@ def __init__(
  features_start: int = 64,
  bilinear: bool = False,
  ):
-
  if num_layers < 1:
  raise ValueError(f"num_layers = {num_layers}, expected: num_layers > 0")
 

@@ -90,14 +90,11 @@ def gather_lit_args(cls: Any, root_cls: Optional[Any] = None) -> List[LitArg]:
  arguments: List[LitArg] = []
  argument_names = []
  for obj in inspect.getmro(cls):
-
  if obj is root_cls and len(arguments) > 0:
  break
 
  if issubclass(obj, root_cls):
-
  default_params = inspect.signature(obj.__init__).parameters # type: ignore
-
  for arg in default_params:
  arg_type = default_params[arg].annotation
  arg_default = default_params[arg].default

@@ -185,7 +185,6 @@ def test_source_is_done_2step_episode(self):
  self.source.histories[0].append(self.exp1)
 
  for idx, exp in enumerate(self.source.runner(self.device)):
-
  self.assertTrue(isinstance(exp, tuple))
 
  if idx == 0:
@@ -211,7 +210,6 @@ def test_source_is_done_metrics(self):
  history += [self.exp1, self.exp2, self.exp2]
 
  for idx, exp in enumerate(self.source.runner(self.device)):
-
  if idx == n_steps - 1:
  self.assertEqual(self.source._total_rewards[0], 1)
  self.assertEqual(self.source.total_steps[0], 1)

@@ -29,7 +29,6 @@
 @pytest.mark.parametrize("batch_size,num_samples", [(16, 100), (1, 0)])
 def test_dummy_ds(catch_warnings, batch_size, num_samples):
  if num_samples > 0:
-
  ds = DummyDataset((1, 28, 28), (1,), num_samples=num_samples)
  dl = DataLoader(ds, batch_size=batch_size)
 

@@ -13,7 +13,6 @@
 
 class TestRLLoss(TestCase):
  def setUp(self) -> None:
-
  self.state = torch.rand(32, 4, 84, 84)
  self.next_state = torch.rand(32, 4, 84, 84)
  self.action = torch.ones([32])

@@ -31,7 +31,6 @@ def setUp(self) -> None:
  self.value_agent = ValueAgent(self.net, self.env.action_space.n)
 
  def test_value_agent(self):
-
  action = self.value_agent(self.state, self.device)
  self.assertIsInstance(action, list)
  self.assertIsInstance(action[0], int)

@@ -7,7 +7,6 @@
 
 
 class DummyParentModel(LightningModule):
-
  name = "parent-model"
 
  def __init__(self, a: int, b: str, c: str = "parent_model_c"):
@@ -19,7 +18,6 @@ def forward(self, x):
 
 
 class DummyParentDataModule(LightningDataModule):
-
  name = "parent-dm"
 
  def __init__(self, d: str, c: str = "parent_dm_c"):