Update versioning system (#252)

* Draft for new versioning scheme

* Update build system

* Testing file

* Fix build error

* Use build

* Python version

* Importlib in Python 3.8

* Add back requirements_test.txt

* Rename

* Bad torch version

* Delete version file

* Ignore version

* No need to build

* Delete test requirements

* Apply isort

* Fix Python version

.github/workflows/build-and-test.yml

@@ -15,7 +15,7 @@ jobs:
       max-parallel: 4
       matrix:
         platform: [ubuntu-latest]
-        python-version: [3.7, 3.8]
+        python-version: [3.8]
 
     runs-on: ${{ matrix.platform }}
 
@@ -28,8 +28,9 @@ jobs:
       - name: Install
         run: |
           python -m pip install --upgrade pip
-          pip install --upgrade wheel
-          pip install -e .
+          pip install --upgrade wheel build setuptools
+          python -m build .
+          pip install dist/*.whl
       - name: Test Import
         run: |
           python -c 'import fastmri'
@@ -40,7 +41,7 @@ jobs:
       max-parallel: 4
       matrix:
         platform: [ubuntu-latest]
-        python-version: [3.7, 3.8]
+        python-version: [3.8]
 
     runs-on: ${{ matrix.platform }}
 
@@ -65,7 +66,7 @@ jobs:
         run: |
           python -m pip install --upgrade pip
           pip install --upgrade wheel
-          pip install -r dev-requirements.txt
+          pip install --editable ".[tests]"
       - name: Check Formatting and Lint
         run: |
           python --version
@@ -75,10 +76,8 @@ jobs:
           mypy fastmri
           flake8 --version
           flake8 fastmri_examples fastmri tests
-      - name: Install fastMRI
-        run: |
-          python setup.py bdist_wheel
-          pip install dist/*.whl
+          isort --version
+          isort --check-only fastmri tests fastmri_examples
       - name: Run pytest
         run: |
           echo -e "PyTorch \c" && pip show torch | grep Version 

.gitignore

@@ -20,3 +20,4 @@ lightning_logs/
 *.pt
 build/
 dist/
+fastmri/_version.py

README.md

@@ -54,7 +54,8 @@ help(SliceDataset)
 
 **Note:** Contributions to the code are continuously tested via GitHub actions.
 If you encounter an issue, the best first thing to do is to try to match the
-test environments in `requirements.txt` and `dev-requirements.txt`.
+`tests` environment in `setup.cfg`, e.g., `pip install --editable ".[tests]"`
+when installing from source.
 
 **Note:** As documented in [Issue 215](https://github.com/facebookresearch/fastMRI/issues/215),
 there is currently a memory leak when using `h5py` installed from `pip` and

fastmri/__init__.py

@@ -5,14 +5,16 @@
 LICENSE file in the root directory of this source tree.
 """
 
-__version__ = "0.1.2a20220121"
-__author__ = "Facebook/NYU fastMRI Team"
-__author_email__ = "[email protected]"
-__license__ = "MIT"
-__homepage__ = "https://fastmri.org/"
+from importlib.metadata import PackageNotFoundError, version
+
+try:
+    __version__ = version("fastmri")
+except PackageNotFoundError:
+    # package is not installed
+    import warnings
+
+    warnings.warn("Could not retrieve fastmri version!")
 
-import torch
-from packaging import version
 
 from .coil_combine import rss, rss_complex
 from .fftc import fft2c_new as fft2c

fastmri/coil_combine.py

@@ -23,7 +23,7 @@ def rss(data: torch.Tensor, dim: int = 0) -> torch.Tensor:
     Returns:
         The RSS value.
     """
-    return torch.sqrt((data ** 2).sum(dim))
+    return torch.sqrt((data**2).sum(dim))
 
 
 def rss_complex(data: torch.Tensor, dim: int = 0) -> torch.Tensor:

fastmri/data/__init__.py

@@ -5,5 +5,5 @@
 LICENSE file in the root directory of this source tree.
 """
 
-from .mri_data import SliceDataset, CombinedSliceDataset
+from .mri_data import CombinedSliceDataset, SliceDataset
 from .volume_sampler import VolumeSampler

fastmri/data/mri_data.py

@@ -276,8 +276,8 @@ def __init__(
             if dataset_cache.get(root) is None and use_dataset_cache:
                 dataset_cache[root] = self.examples
                 logging.info(f"Saving dataset cache to {self.dataset_cache_file}.")
-                with open(self.dataset_cache_file, "wb") as f:
-                    pickle.dump(dataset_cache, f)
+                with open(self.dataset_cache_file, "wb") as cache_f:
+                    pickle.dump(dataset_cache, cache_f)
         else:
             logging.info(f"Using dataset cache from {self.dataset_cache_file}.")
             self.examples = dataset_cache[root]

fastmri/data/transforms.py

@@ -7,10 +7,11 @@
 
 from typing import Dict, NamedTuple, Optional, Sequence, Tuple, Union
 
-import fastmri
 import numpy as np
 import torch
 
+import fastmri
+
 from .subsample import MaskFunc
 
 

fastmri/data/volume_sampler.py

@@ -9,9 +9,10 @@
 
 import torch
 import torch.distributed as dist
-from fastmri.data.mri_data import CombinedSliceDataset, SliceDataset
 from torch.utils.data import Sampler
 
+from fastmri.data.mri_data import CombinedSliceDataset, SliceDataset
+
 
 class VolumeSampler(Sampler):
     """

fastmri/evaluate.py

@@ -25,7 +25,7 @@ def mse(gt: np.ndarray, pred: np.ndarray) -> np.ndarray:
 
 def nmse(gt: np.ndarray, pred: np.ndarray) -> np.ndarray:
     """Compute Normalized Mean Squared Error (NMSE)"""
-    return np.linalg.norm(gt - pred) ** 2 / np.linalg.norm(gt) ** 2
+    return np.array(np.linalg.norm(gt - pred) ** 2 / np.linalg.norm(gt) ** 2)
 
 
 def psnr(

fastmri/losses.py

@@ -25,8 +25,8 @@ def __init__(self, win_size: int = 7, k1: float = 0.01, k2: float = 0.03):
         super().__init__()
         self.win_size = win_size
         self.k1, self.k2 = k1, k2
-        self.register_buffer("w", torch.ones(1, 1, win_size, win_size) / win_size ** 2)
-        NP = win_size ** 2
+        self.register_buffer("w", torch.ones(1, 1, win_size, win_size) / win_size**2)
+        NP = win_size**2
         self.cov_norm = NP / (NP - 1)
 
     def forward(
@@ -52,7 +52,7 @@ def forward(
         A1, A2, B1, B2 = (
             2 * ux * uy + C1,
             2 * vxy + C2,
-            ux ** 2 + uy ** 2 + C1,
+            ux**2 + uy**2 + C1,
             vx + vy + C2,
         )
         D = B1 * B2

fastmri/math.py

@@ -66,7 +66,7 @@ def complex_abs(data: torch.Tensor) -> torch.Tensor:
     if not data.shape[-1] == 2:
         raise ValueError("Tensor does not have separate complex dim.")
 
-    return (data ** 2).sum(dim=-1).sqrt()
+    return (data**2).sum(dim=-1).sqrt()
 
 
 def complex_abs_sq(data: torch.Tensor) -> torch.Tensor:
@@ -83,7 +83,7 @@ def complex_abs_sq(data: torch.Tensor) -> torch.Tensor:
     if not data.shape[-1] == 2:
         raise ValueError("Tensor does not have separate complex dim.")
 
-    return (data ** 2).sum(dim=-1)
+    return (data**2).sum(dim=-1)
 
 
 def tensor_to_complex_np(data: torch.Tensor) -> np.ndarray:

fastmri/models/policy.py

@@ -7,8 +7,7 @@
 
 import functools
 import operator
-from typing import List
-from typing import Tuple
+from typing import List, Tuple
 
 import torch
 import torch.nn as nn
@@ -64,14 +63,10 @@ def forward(self, mask: torch.Tensor, kspace: torch.Tensor):
         if self.use_softplus:
             # Softplus to make positive
             prob_mask = F.softplus(sampler_out, beta=self.slope)
-            prob_mask = (
-                prob_mask
-                / torch.max(
-                    (1 - mask.reshape(prob_mask.shape[0], prob_mask.shape[1]))
-                    * prob_mask,
-                    dim=1,
-                )[0].reshape(-1, 1)
-            )
+            prob_mask = prob_mask / torch.max(
+                (1 - mask.reshape(prob_mask.shape[0], prob_mask.shape[1])) * prob_mask,
+                dim=1,
+            )[0].reshape(-1, 1)
         else:
             # Sigmoid to make positive
             prob_mask = torch.sigmoid(self.slope * sampler_out)
@@ -419,7 +414,7 @@ def __init__(
         self.down_sample_layers = nn.ModuleList(
             [
                 SingleConvBlock(
-                    chans * 2 ** i,
+                    chans * 2**i,
                     chans * 2 ** (i + 1),
                     drop_prob,
                     pool_size=self.pool_size,

fastmri/models/varnet.py

@@ -6,12 +6,13 @@
 """
 
 import math
-from typing import List, Tuple, Optional
+from typing import List, Optional, Tuple
 
-import fastmri
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
+
+import fastmri
 from fastmri.data import transforms
 
 from .unet import Unet

fastmri/pl_modules/__init__.py

@@ -5,7 +5,7 @@
 LICENSE file in the root directory of this source tree.
 """
 
+from .data_module import FastMriDataModule
 from .mri_module import MriModule
 from .unet_module import UnetModule
 from .varnet_module import VarNetModule
-from .data_module import FastMriDataModule

fastmri/pl_modules/data_module.py

@@ -9,9 +9,10 @@
 from pathlib import Path
 from typing import Callable, Optional, Union
 
-import fastmri
 import pytorch_lightning as pl
 import torch
+
+import fastmri
 from fastmri.data import CombinedSliceDataset, SliceDataset
 
 
@@ -52,13 +53,13 @@ def worker_init_fn(worker_id):
                         + worker_info.id * len(data.datasets)
                         + i
                     )
-                dataset.transform.mask_func.rng.seed(seed_i % (2 ** 32 - 1))
+                dataset.transform.mask_func.rng.seed(seed_i % (2**32 - 1))
     elif data.transform.mask_func is not None:
         if is_ddp:  # DDP training: unique seed is determined by worker and device
             seed = base_seed + torch.distributed.get_rank() * worker_info.num_workers
         else:
             seed = base_seed
-        data.transform.mask_func.rng.seed(seed % (2 ** 32 - 1))
+        data.transform.mask_func.rng.seed(seed % (2**32 - 1))
 
 
 def _check_both_not_none(val1, val2):

fastmri/pl_modules/mri_module.py

@@ -9,13 +9,14 @@
 from argparse import ArgumentParser
 from collections import defaultdict
 
-import fastmri
 import numpy as np
 import pytorch_lightning as pl
 import torch
-from fastmri import evaluate
 from torchmetrics.metric import Metric
 
+import fastmri
+from fastmri import evaluate
+
 
 class DistributedMetricSum(Metric):
     def __init__(self, dist_sync_on_step=True):

fastmri/pl_modules/unet_module.py

@@ -8,9 +8,10 @@
 from argparse import ArgumentParser
 
 import torch
-from fastmri.models import Unet
 from torch.nn import functional as F
 
+from fastmri.models import Unet
+
 from .mri_module import MriModule
 
 

fastmri/pl_modules/varnet_module.py

@@ -7,8 +7,9 @@
 
 from argparse import ArgumentParser
 
-import fastmri
 import torch
+
+import fastmri
 from fastmri.data import transforms
 from fastmri.models import VarNet
 

fastmri_examples/adaptive_varnet/eval_pretrained_adaptive_varnet.py

@@ -4,14 +4,14 @@
 import numpy as np
 import pytorch_lightning as pl
 import torch
+from pl_modules import AdaptiveVarNetModule, VarNetModule
+from subsample import create_mask_for_mask_type
+
 from fastmri import evaluate
 from fastmri.data.mri_data import fetch_dir
 from fastmri.data.transforms import MiniCoilTransform
 from fastmri.pl_modules import FastMriDataModule
 
-from pl_modules import AdaptiveVarNetModule, VarNetModule
-from subsample import create_mask_for_mask_type
-
 
 def str2bool(v):
     if isinstance(v, bool):

fastmri_examples/adaptive_varnet/pl_modules/adaptive_varnet_module.py

@@ -9,9 +9,10 @@
 from collections import defaultdict
 from typing import Tuple
 
-import fastmri
 import numpy as np
 import torch
+
+import fastmri
 from fastmri import evaluate
 from fastmri.data import transforms
 from fastmri.models import AdaptiveVarNet

fastmri_examples/adaptive_varnet/pl_modules/varnet_module.py

@@ -9,18 +9,19 @@
 from collections import defaultdict
 from typing import Optional
 
-import fastmri
 import numpy as np
 import torch
 import torch.nn as nn
+
+import fastmri
 from fastmri import evaluate
 from fastmri.data import transforms
 from fastmri.data.transforms import VarNetSample
 from fastmri.models.adaptive_varnet import AdaptiveSensitivityModel, AdaptiveVarNetBlock
 from fastmri.models.varnet import NormUnet
 from fastmri.pl_modules.mri_module import MriModule
 
-from .metrics import DistributedMetricSum, DistributedArraySum
+from .metrics import DistributedArraySum, DistributedMetricSum
 
 
 class VarNet(nn.Module):

fastmri_examples/adaptive_varnet/subsample.py

@@ -8,6 +8,7 @@
 from typing import Optional, Sequence
 
 import numpy as np
+
 from fastmri.data.subsample import MaskFunc, RandomMaskFunc