Merge branch 'Lightning-AI:master' into master

.azure/gpu-tests.yml

@@ -25,7 +25,7 @@ jobs:
  container:
  image: "pytorchlightning/pytorch_lightning:base-cuda-py3.9-torch1.12-cuda11.6.1"
  #endpoint: azureContainerRegistryConnection
- options: "--gpus=all --shm-size=16g --name ci-container -v /usr/bin/docker:/tmp/docker:ro"
+ options: "--gpus=all --shm-size=16g -v /usr/bin/docker:/tmp/docker:ro"
 
  workspace:
  clean: all
@@ -34,12 +34,14 @@ jobs:
 
  - bash: |
  echo "##vso[task.setvariable variable=CUDA_VISIBLE_DEVICES]$(DEVICES)"
- displayName: 'set visible devices'
+ echo "##vso[task.setvariable variable=CONTAINER_ID]$(head -1 /proc/self/cgroup|cut -d/ -f3)"
+ displayName: 'Set environment variables'
 
  - bash: |
  whereis nvidia
  nvidia-smi
  echo $CUDA_VISIBLE_DEVICES
+ echo $CONTAINER_ID
  python --version
  pip --version
  pip list
@@ -50,7 +52,7 @@ jobs:
  displayName: 'Sanity check'
 
  - script: |
- /tmp/docker exec -t -u 0 ci-container \
+ /tmp/docker exec -t -u 0 $CONTAINER_ID \
  sh -c "apt-get update && DEBIAN_FRONTEND=noninteractive apt-get -o Dpkg::Options::="--force-confold" -y install sudo"
  displayName: 'Install Sudo in container (thanks Microsoft!)'
 

.github/workflows/ci-tests.yml

@@ -69,12 +69,9 @@ jobs:
  --ignore=pl_bolts/models/self_supervised/amdim/transforms.py \
  --ignore=pl_bolts/models/rl
 
- # Note: This uses an internal pip API and may not always work
- # https://github.com/actions/cache/blob/master/examples.md#multiple-oss-in-a-workflow
- - name: Get pip cache
+ - name: Get pip cache dir
  id: pip-cache
- run: |
- python -c "from pip._internal.locations import USER_CACHE_DIR; print('::set-output name=dir::' + USER_CACHE_DIR)"
+ run: echo "dir=$(pip cache dir)" >> $GITHUB_OUTPUT
 
  - name: Cache pip
  uses: actions/cache@v3

pl_bolts/models/gans/dcgan/components.py

@@ -1,10 +1,8 @@
 # Based on https://github.com/pytorch/examples/blob/master/dcgan/main.py
-from torch import Tensor, nn
+import torch.nn as nn
+from torch import Tensor
 
-from pl_bolts.utils.stability import under_review
 
-
-@under_review()
 class DCGANGenerator(nn.Module):
  def __init__(self, latent_dim: int, feature_maps: int, image_channels: int) -> None:
  """
@@ -50,7 +48,6 @@ def forward(self, noise: Tensor) -> Tensor:
  return self.gen(noise)
 
 
-@under_review()
 class DCGANDiscriminator(nn.Module):
  def __init__(self, feature_maps: int, image_channels: int) -> None:
  """

pl_bolts/models/gans/dcgan/dcgan_module.py

@@ -2,14 +2,14 @@
 from typing import Any
 
 import torch
+import torch.nn as nn
 from pytorch_lightning import LightningModule, Trainer, seed_everything
-from torch import Tensor, nn
+from torch import Tensor
 from torch.utils.data import DataLoader
 
 from pl_bolts.callbacks import LatentDimInterpolator, TensorboardGenerativeModelImageSampler
 from pl_bolts.models.gans.dcgan.components import DCGANDiscriminator, DCGANGenerator
 from pl_bolts.utils import _TORCHVISION_AVAILABLE
-from pl_bolts.utils.stability import under_review
 from pl_bolts.utils.warnings import warn_missing_pkg
 
 if _TORCHVISION_AVAILABLE:
@@ -19,7 +19,6 @@
  warn_missing_pkg("torchvision")
 
 
-@under_review()
 class DCGAN(LightningModule):
  """DCGAN implementation.
 
@@ -28,7 +27,7 @@ class DCGAN(LightningModule):
  from pl_bolts.models.gans import DCGAN
 
  m = DCGAN()
- Trainer(gpus=2).fit(m)
+ Trainer(accelerator="gpu", devices=2).fit(m)
 
  Example CLI::
 
@@ -80,10 +79,10 @@ def _get_discriminator(self) -> nn.Module:
  def _weights_init(m):
  classname = m.__class__.__name__
  if classname.find("Conv") != -1:
- torch.nn.init.normal_(m.weight, 0.0, 0.02)
+ nn.init.normal_(m.weight, 0.0, 0.02)
  elif classname.find("BatchNorm") != -1:
- torch.nn.init.normal_(m.weight, 1.0, 0.02)
- torch.nn.init.zeros_(m.bias)
+ nn.init.normal_(m.weight, 1.0, 0.02)
+ nn.init.zeros_(m.bias)
 
  def configure_optimizers(self):
  lr = self.hparams.learning_rate
@@ -173,7 +172,6 @@ def add_model_specific_args(parent_parser: ArgumentParser) -> ArgumentParser:
  return parser
 
 
-@under_review()
 def cli_main(args=None):
  seed_everything(1234)
 

pl_bolts/utils/__init__.py

@@ -21,5 +21,7 @@
 _TORCH_ORT_AVAILABLE = module_available("torch_ort")
 _TORCH_MAX_VERSION_SPARSEML = compare_version("torch", operator.lt, "1.11.0")
 _SPARSEML_AVAILABLE = module_available("sparseml") and _PL_GREATER_EQUAL_1_4_5 and _TORCH_MAX_VERSION_SPARSEML
+_JSONARGPARSE_GREATER_THAN_4_16_0 = compare_version("jsonargparse", operator.gt, "4.16.0")
+
 
 __all__ = ["BatchGradientVerification"]

requirements.txt

@@ -1,5 +1,5 @@
 torch>=1.9.*
 torchmetrics>=0.4.1
 pytorch-lightning>=1.6.0
-lightning-utilities==0.3.0
+lightning-utilities==0.4.*, !=0.4.0
 packaging

tests/models/test_scripts.py

@@ -2,6 +2,7 @@
 
 import pytest
 
+from pl_bolts.utils import _JSONARGPARSE_GREATER_THAN_4_16_0
 from tests import _MARK_REQUIRE_GPU, DATASETS_PATH
 
 _DEFAULT_ARGS = f" --data_dir {DATASETS_PATH} --max_epochs 1 --max_steps 2 --batch_size 4"
@@ -13,9 +14,7 @@
 @pytest.mark.parametrize("dataset_name", ["mnist", "cifar10"])
 @pytest.mark.parametrize(
  "cli_args",
- [
- " --dataset %(dataset_name)s" + _DEFAULT_ARGS,
- ],
+ [" --dataset %(dataset_name)s" + _DEFAULT_ARGS],
 )
 def test_cli_run_basic_gan(cli_args, dataset_name):
  from pl_bolts.models.gans.basic.basic_gan_module import cli_main
@@ -112,6 +111,9 @@ def test_cli_run_vision_image_gpt(cli_args):
 
 @pytest.mark.parametrize("cli_args", [_DEFAULT_LIGHTNING_CLI_ARGS + " --trainer.gpus 1"])
 @pytest.mark.skipif(**_MARK_REQUIRE_GPU)
+@pytest.mark.skipif(
+ not _JSONARGPARSE_GREATER_THAN_4_16_0, reason="Failing on CI, need to be fixed"
+) # see https://github.com/omni-us/jsonargparse/issues/187
 def test_cli_run_retinanet(cli_args):
  """Test running CLI for an example with default params."""
  from pl_bolts.models.detection.retinanet.retinanet_module import cli_main