upgrade pytorch lightning version (#458)

* upgrade lightning version

.flake8

@@ -2,7 +2,7 @@
 # Keep in sync with setup.cfg which is used for source packages.
 
 [flake8]
-ignore = W503, E203, E221, C901, C408, E741, C407, E741, B006, B007, B017, B950, C416
+ignore = W503, E203, E221, C901, C408, E741, C407, E741, B006, B007, B017, B950, C416, E203
 max-line-length = 88
 max-complexity = 18
 select = B,C,E,F,W,T4,B9

.github/workflows/workflow.yml

@@ -11,15 +11,15 @@ jobs:
  runs-on: ubuntu-latest
  steps:
  - uses: actions/checkout@v2
- - name: Set up Python 3.6
+ - name: Set up Python 3.8
  uses: actions/setup-python@v2
  with:
- python-version: 3.6
+ python-version: 3.8
  - name: Install dependencies
  run: |
  python -m pip install --upgrade pip
- python -m pip install flake8==3.9.2 flake8-bugbear flake8-comprehensions isort==5.8.0
- python -m pip install black==21.6b0
+ python -m pip install flake8==5.0.4 isort==5.10.1
+ python -m pip install black==22.6.0
  flake8 --version
  - name: Lint
  run: |
@@ -34,21 +34,23 @@ jobs:
  runs-on: ubuntu-latest
  strategy:
  matrix:
- torch: [1.7.0, 1.8.1, 1.9.0]
+ torch: [1.9.0, 1.10.1, 1.11.0, 1.12.1]
  include:
- - torch: 1.7.0
- torchvision: 0.8.1
- - torch: 1.8.1
- torchvision: 0.9.1
  - torch: 1.9.0
  torchvision: 0.10.0
+ - torch: 1.10.1
+ torchvision: 0.11.2
+ - torch: 1.11.0
+ torchvision: 0.12.0
+ - torch: 1.12.1
+ torchvision: 0.13.1
  steps:
  - name: Checkout
  uses: actions/checkout@v2
- - name: Set up Python 3.6
+ - name: Set up Python 3.8
  uses: actions/setup-python@v2
  with:
- python-version: 3.6
+ python-version: 3.8
  - name: Install dependencies
  run: |
  python -m pip install -U pip
@@ -65,7 +67,7 @@ jobs:
  coverage report -m
  - name: Upload coverage to Codecov
  uses: codecov/codecov-action@v2
- if: matrix.torch == '1.9.0'
+ if: matrix.torch == '1.12.1'
  with:
  file: ./coverage.xml
  flags: unittests

.pre-commit-config.yaml

@@ -10,17 +10,17 @@ repos:
  - id: requirements-txt-fixer
 
  - repo: https://github.com/pycqa/isort
- rev: 5.8.0
+ rev: 5.10.1
  hooks:
  - id: isort
  args: ["--profile", "black"]
 
  - repo: https://github.com/psf/black
- rev: 21.6b0
+ rev: 22.6.0
  hooks:
  - id: black
 
  - repo: https://gitlab.com/pycqa/flake8
- rev: 3.9.2
+ rev: 5.0.4
  hooks:
  - id: flake8

README.md

@@ -73,6 +73,8 @@ MobileDet | 320*320 | 25.6 | - | -
 
 ## NEWS!!!
 
+* [2022.08.26] Upgrade to [pytorch-lightning-1.7](https://lightning.ai/). The minimum PyTorch version is upgraded to 1.9. To use previous version of PyTorch, please install [NanoDet <= v1.0.0-alpha-1](https://github.com/RangiLyu/nanodet/tags)
+
 * [2021.12.25] **NanoDet-Plus** release! Adding **AGM**(Assign Guidance Module) & **DSLA**(Dynamic Soft Label Assigner) to improve **7 mAP** with only a little cost.
 
 Find more update notes in [Update notes](docs/update.md).
@@ -139,7 +141,7 @@ Besides, We provide a notebook [here](./demo/demo-inference-with-pytorch.ipynb)
 * Linux or MacOS
 * CUDA >= 10.0
 * Python >= 3.6
-* Pytorch >= 1.7
+* Pytorch >= 1.9
 * experimental support Windows (Notice: Windows not support distributed training before pytorch1.7)
 
 ### Step
@@ -169,7 +171,7 @@ cd nanodet
 ```shell script
 pip install -r requirements.txt
 ```
- 
+
 5. Setup NanoDet
 ```shell script
 python setup.py develop

nanodet/model/loss/iou_loss.py

@@ -271,7 +271,7 @@ def diou_loss(pred, target, eps=1e-7):
  cw = enclose_wh[:, 0]
  ch = enclose_wh[:, 1]
 
- c2 = cw ** 2 + ch ** 2 + eps
+ c2 = cw**2 + ch**2 + eps
 
  b1_x1, b1_y1 = pred[:, 0], pred[:, 1]
  b1_x2, b1_y2 = pred[:, 2], pred[:, 3]
@@ -326,7 +326,7 @@ def ciou_loss(pred, target, eps=1e-7):
  cw = enclose_wh[:, 0]
  ch = enclose_wh[:, 1]
 
- c2 = cw ** 2 + ch ** 2 + eps
+ c2 = cw**2 + ch**2 + eps
 
  b1_x1, b1_y1 = pred[:, 0], pred[:, 1]
  b1_x2, b1_y2 = pred[:, 2], pred[:, 3]
@@ -340,11 +340,11 @@ def ciou_loss(pred, target, eps=1e-7):
  right = ((b2_y1 + b2_y2) - (b1_y1 + b1_y2)) ** 2 / 4
  rho2 = left + right
 
- factor = 4 / math.pi ** 2
+ factor = 4 / math.pi**2
  v = factor * torch.pow(torch.atan(w2 / h2) - torch.atan(w1 / h1), 2)
 
  # CIoU
- cious = ious - (rho2 / c2 + v ** 2 / (1 - ious + v))
+ cious = ious - (rho2 / c2 + v**2 / (1 - ious + v))
  loss = 1 - cious
  return loss
 

nanodet/trainer/task.py

@@ -316,7 +316,7 @@ def on_train_start(self) -> None:
  if self.current_epoch > 0:
  self.lr_scheduler.last_epoch = self.current_epoch - 1
 
- def on_pretrain_routine_end(self) -> None:
+ def on_fit_start(self) -> None:
  if "weight_averager" in self.cfg.model:
  self.logger.info("Weight Averaging is enabled")
  if self.weight_averager and self.weight_averager.has_inited():
@@ -327,10 +327,10 @@ def on_pretrain_routine_end(self) -> None:
  )
  self.weight_averager.load_from(self.model)
 
- def on_epoch_start(self):
+ def on_train_epoch_start(self):
  self.model.set_epoch(self.current_epoch)
 
- def on_train_batch_end(self, outputs, batch, batch_idx, dataloader_idx) -> None:
+ def on_train_batch_end(self, outputs, batch, batch_idx) -> None:
  if self.weight_averager:
  self.weight_averager.update(self.model, self.global_step)
 

nanodet/util/flops_counter.py

@@ -132,21 +132,21 @@ def flops_to_string(flops, units="GFLOPs", precision=2):
  '3e-09 FLOPs'
  """
  if units is None:
- if flops // 10 ** 9 > 0:
- return str(round(flops / 10.0 ** 9, precision)) + " GFLOPs"
- elif flops // 10 ** 6 > 0:
- return str(round(flops / 10.0 ** 6, precision)) + " MFLOPs"
- elif flops // 10 ** 3 > 0:
- return str(round(flops / 10.0 ** 3, precision)) + " KFLOPs"
+ if flops // 10**9 > 0:
+ return str(round(flops / 10.0**9, precision)) + " GFLOPs"
+ elif flops // 10**6 > 0:
+ return str(round(flops / 10.0**6, precision)) + " MFLOPs"
+ elif flops // 10**3 > 0:
+ return str(round(flops / 10.0**3, precision)) + " KFLOPs"
  else:
  return str(flops) + " FLOPs"
  else:
  if units == "GFLOPs":
- return str(round(flops / 10.0 ** 9, precision)) + " " + units
+ return str(round(flops / 10.0**9, precision)) + " " + units
  elif units == "MFLOPs":
- return str(round(flops / 10.0 ** 6, precision)) + " " + units
+ return str(round(flops / 10.0**6, precision)) + " " + units
  elif units == "KFLOPs":
- return str(round(flops / 10.0 ** 3, precision)) + " " + units
+ return str(round(flops / 10.0**3, precision)) + " " + units
  else:
  return str(flops) + " FLOPs"
 
@@ -170,17 +170,17 @@ def params_to_string(num_params, units=None, precision=2):
  '3e-09'
  """
  if units is None:
- if num_params // 10 ** 6 > 0:
- return str(round(num_params / 10 ** 6, precision)) + " M"
- elif num_params // 10 ** 3:
- return str(round(num_params / 10 ** 3, precision)) + " k"
+ if num_params // 10**6 > 0:
+ return str(round(num_params / 10**6, precision)) + " M"
+ elif num_params // 10**3:
+ return str(round(num_params / 10**3, precision)) + " k"
  else:
  return str(num_params)
  else:
  if units == "M":
- return str(round(num_params / 10.0 ** 6, precision)) + " " + units
+ return str(round(num_params / 10.0**6, precision)) + " " + units
  elif units == "K":
- return str(round(num_params / 10.0 ** 3, precision)) + " " + units
+ return str(round(num_params / 10.0**3, precision)) + " " + units
  else:
  return str(num_params)
 

nanodet/util/logger.py

@@ -17,8 +17,8 @@
 import time
 
 import numpy as np
-from pytorch_lightning.loggers import LightningLoggerBase
-from pytorch_lightning.loggers.base import rank_zero_experiment
+from pytorch_lightning.loggers import Logger as LightningLoggerBase
+from pytorch_lightning.loggers.logger import rank_zero_experiment
 from pytorch_lightning.utilities import rank_zero_only
 from pytorch_lightning.utilities.cloud_io import get_filesystem
 from termcolor import colored

requirements.txt

@@ -7,11 +7,11 @@ onnx-simplifier
 opencv-python
 pyaml
 pycocotools
-pytorch-lightning>=1.4.0
+pytorch-lightning>=1.7.0
 tabulate
 tensorboard
 termcolor
-torch>=1.7
+torch>=1.9
 torchmetrics
 torchvision
 tqdm

tests/test_models/test_fpn/test_fpn.py

@@ -8,7 +8,7 @@ def test_fpn():
  """Tests fpn."""
  s = 64
  in_channels = [8, 16, 32, 64]
- feat_sizes = [s // 2 ** i for i in range(4)] # [64, 32, 16, 8]
+ feat_sizes = [s // 2**i for i in range(4)] # [64, 32, 16, 8]
  out_channels = 8
  # `num_outs` is not equal to len(in_channels) - start_level
  with pytest.raises(AssertionError):

tests/test_models/test_fpn/test_ghost_pan.py

@@ -8,7 +8,7 @@ def test_ghost_pan():
  """Tests GhostPAN."""
  s = 64
  in_channels = [8, 16, 32, 64]
- feat_sizes = [s // 2 ** i for i in range(4)] # [64, 32, 16, 8]
+ feat_sizes = [s // 2**i for i in range(4)] # [64, 32, 16, 8]
  out_channels = 8
  # `num_extra_level` >= 0
  with pytest.raises(AssertionError):
@@ -34,7 +34,7 @@ def test_ghost_pan():
  assert len(outs) == len(in_channels) + 1
  for i in range(len(in_channels)):
  assert outs[i].shape[1] == out_channels
- assert outs[i].shape[2] == outs[i].shape[3] == s // (2 ** i)
+ assert outs[i].shape[2] == outs[i].shape[3] == s // (2**i)
  assert outs[-1].shape[1] == out_channels
  assert outs[-1].shape[2] == outs[-1].shape[3] == s // 2 ** len(in_channels)
 
@@ -61,6 +61,6 @@ def test_ghost_pan():
  assert len(outs) == len(in_channels) + 1
  for i in range(len(in_channels)):
  assert outs[i].shape[1] == out_channels
- assert outs[i].shape[2] == outs[i].shape[3] == s // (2 ** i)
+ assert outs[i].shape[2] == outs[i].shape[3] == s // (2**i)
  assert outs[-1].shape[1] == out_channels
  assert outs[-1].shape[2] == outs[-1].shape[3] == s // 2 ** len(in_channels)

tests/test_models/test_fpn/test_pan.py

@@ -8,7 +8,7 @@ def test_pan():
  """Tests PAN."""
  s = 64
  in_channels = [8, 16, 32, 64]
- feat_sizes = [s // 2 ** i for i in range(4)] # [64, 32, 16, 8]
+ feat_sizes = [s // 2**i for i in range(4)] # [64, 32, 16, 8]
  out_channels = 8
  # `num_outs` is not equal to len(in_channels) - start_level
  with pytest.raises(AssertionError):

tests/test_models/test_fpn/test_tan.py

@@ -8,7 +8,7 @@ def test_tan():
  """Tests TAN."""
  s = 64
  in_channels = [8, 16, 32]
- feat_sizes = [s // 2 ** i for i in range(3)] # [64, 32, 16]
+ feat_sizes = [s // 2**i for i in range(3)] # [64, 32, 16]
  out_channels = 8
 
  with pytest.raises(AssertionError):
@@ -41,4 +41,4 @@ def test_tan():
  assert len(outs) == 3
  for i in range(3):
  assert outs[i].shape[1] == out_channels
- assert outs[i].shape[2] == outs[i].shape[3] == s // (2 ** i)
+ assert outs[i].shape[2] == outs[i].shape[3] == s // (2**i)

tests/test_models/test_modules/test_norm.py

@@ -46,11 +46,11 @@ def test_build_norm_layer():
  cfg = dict(type=type_name)
  if type_name == "GN":
  cfg["num_groups"] = 2
- name, layer = build_norm_layer(cfg, 3, postfix=postfix)
+ name, layer = build_norm_layer(cfg, 4, postfix=postfix)
  assert name == abbr_mapping[type_name] + str(postfix)
  assert isinstance(layer, module)
  if type_name == "GN":
- assert layer.num_channels == 3
+ assert layer.num_channels == 4
  assert layer.num_groups == cfg["num_groups"]
  elif type_name != "LN":
- assert layer.num_features == 3
+ assert layer.num_features == 4

tests/test_trainer/test_lightning_task.py

@@ -13,7 +13,7 @@ class DummyTrainer(nn.Module):
  global_step = 0
  local_rank = 0
  use_ddp = False
- logger = NanoDetLightningLogger(tempfile.TemporaryDirectory().name)
+ loggers = [NanoDetLightningLogger(tempfile.TemporaryDirectory().name)]
 
  def save_checkpoint(self, *args, **kwargs):
  pass
@@ -24,7 +24,7 @@ def __init__(self, task):
  self.task = task
 
  def test(self):
- self.task.trainer = DummyTrainer()
+ self.task._trainer = DummyTrainer()
 
  optimizer = self.task.configure_optimizers()
 

tools/test.py

@@ -83,10 +83,16 @@ def main(args):
  ckpt = convert_old_model(ckpt)
  task.load_state_dict(ckpt["state_dict"])
 
+ if cfg.device.gpu_ids == -1:
+ logger.info("Using CPU training")
+ accelerator, devices = "cpu", None
+ else:
+ accelerator, devices = "gpu", cfg.device.gpu_ids
+
  trainer = pl.Trainer(
  default_root_dir=cfg.save_dir,
- gpus=cfg.device.gpu_ids,
- accelerator="ddp",
+ accelerator=accelerator,
+ devices=devices,
  log_every_n_steps=cfg.log.interval,
  num_sanity_val_steps=0,
  logger=logger,