0.2.0 release (#348)

* 0.2.0 release

* Fix notebooks

* Fix minimal dependency tests

* Fix integration tests

* Fix integration tests

* Try to avoid running GitHub Actions twice on release PRs

* Revert "Try to avoid running GitHub Actions twice on release PRs"

This reverts commit a1ac7ab.

* GeoDatasets use intersection, not addition

* Adding stack_samples to benchmarks

* Fix zero division error in SEN12MS tests

* Replaces test models with dummy models (#344)

Co-authored-by: Adam J. Stewart <[email protected]>

* lc values must be < num_classes

* updated indices tutorial with latest indices

Co-authored-by: Caleb Robinson <[email protected]>
Co-authored-by: isaaccorley <[email protected]>

docs/tutorials/benchmarking.ipynb

@@ -58,7 +58,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 1,
    "metadata": {
     "gather": {
      "logged": 1629238744113
@@ -75,7 +75,7 @@
     "from torch.utils.data import DataLoader\n",
     "\n",
     "from torchgeo.datasets import NAIP, ChesapeakeDE\n",
-    "from torchgeo.datasets.utils import download_url\n",
+    "from torchgeo.datasets.utils import download_url, stack_samples\n",
     "from torchgeo.samplers import RandomGeoSampler, GridGeoSampler, RandomBatchGeoSampler"
    ]
   },
@@ -209,9 +209,9 @@
     "for cache in [False, True]:\n",
     "    chesapeake = ChesapeakeDE(chesapeake_root, cache=cache)\n",
     "    naip = NAIP(naip_root, crs=chesapeake.crs, res=chesapeake.res, cache=cache)\n",
-    "    dataset = chesapeake + naip\n",
+    "    dataset = chesapeake & naip\n",
     "    sampler = RandomGeoSampler(naip, size=1000, length=888)\n",
-    "    dataloader = DataLoader(dataset, batch_size=12, sampler=sampler)\n",
+    "    dataloader = DataLoader(dataset, batch_size=12, sampler=sampler, collate_fn=stack_samples)\n",
     "    duration, count = time_epoch(dataloader)\n",
     "    print(duration, count)"
    ]
@@ -260,9 +260,9 @@
     "for cache in [False, True]:\n",
     "    chesapeake = ChesapeakeDE(chesapeake_root, cache=cache)\n",
     "    naip = NAIP(naip_root, crs=chesapeake.crs, res=chesapeake.res, cache=cache)\n",
-    "    dataset = chesapeake + naip\n",
+    "    dataset = chesapeake & naip\n",
     "    sampler = GridGeoSampler(naip, size=1000, stride=500)\n",
-    "    dataloader = DataLoader(dataset, batch_size=12, sampler=sampler)\n",
+    "    dataloader = DataLoader(dataset, batch_size=12, sampler=sampler, collate_fn=stack_samples)\n",
     "    duration, count = time_epoch(dataloader)\n",
     "    print(duration, count)"
    ]
@@ -311,9 +311,9 @@
     "for cache in [False, True]:\n",
     "    chesapeake = ChesapeakeDE(chesapeake_root, cache=cache)\n",
     "    naip = NAIP(naip_root, crs=chesapeake.crs, res=chesapeake.res, cache=cache)\n",
-    "    dataset = chesapeake + naip\n",
+    "    dataset = chesapeake & naip\n",
     "    sampler = RandomBatchGeoSampler(naip, size=1000, batch_size=12, length=888)\n",
-    "    dataloader = DataLoader(dataset, batch_sampler=sampler)\n",
+    "    dataloader = DataLoader(dataset, batch_sampler=sampler, collate_fn=stack_samples)\n",
     "    duration, count = time_epoch(dataloader)\n",
     "    print(duration, count)"
    ]
@@ -345,6 +345,7 @@
    "mimetype": "text/x-python",
    "name": "python",
    "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
    "version": "3.9.7"
   },
   "nteract": {

docs/tutorials/trainers.ipynb

@@ -78,7 +78,7 @@
     "import pytorch_lightning as pl\n",
     "from pytorch_lightning.callbacks import EarlyStopping, ModelCheckpoint\n",
     "from pytorch_lightning.loggers import CSVLogger\n",
-    "from torchgeo.datasets import CycloneDataModule\n",
+    "from torchgeo.datamodules import CycloneDataModule\n",
     "from torchgeo.trainers import RegressionTask"
    ]
   },

tests/data/sen12ms/data.py

@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+import hashlib
+import os
+import shutil
+
+import numpy as np
+import rasterio
+from rasterio.transform import Affine
+from rasterio.crs import CRS
+
+SIZE = 64  # image width/height
+
+np.random.seed(0)
+
+meta = {
+    "driver": "GTiff",
+    "nodata": None,
+    "width": SIZE,
+    "height": SIZE,
+    "crs": CRS.from_epsg(32737),
+    "transform": Affine(10.0, 0.0, 390772.3389928384, 0.0, -10.0, 8114182.836060452),
+}
+count = {"lc": 4, "s1": 2, "s2": 13}
+dtype = {"lc": np.uint16, "s1": np.float32, "s2": np.uint16}
+stop = {"lc": 11, "s1": np.iinfo(np.uint16).max, "s2": np.iinfo(np.uint16).max}
+
+file_list = []
+seasons = ["ROIs1158_spring", "ROIs1868_summer", "ROIs1970_fall", "ROIs2017_winter"]
+for season in seasons:
+    # Remove old data
+    if os.path.exists(season):
+        shutil.rmtree(season)
+
+    for source in ["lc", "s1", "s2"]:
+        tarball = f"{season}_{source}.tar.gz"
+
+        # Remove old data
+        if os.path.exists(tarball):
+            os.remove(tarball)
+
+        directory = os.path.join(season, f"{source}_1")
+        os.makedirs(directory)
+
+        # Random images
+        for i in range(1, 3):
+            filename = f"{season}_{source}_1_p{i}.tif"
+            meta["count"] = count[source]
+            meta["dtype"] = dtype[source]
+            with rasterio.open(os.path.join(directory, filename), "w", **meta) as f:
+                for j in range(1, count[source] + 1):
+                    data = np.random.randint(stop[source], size=(SIZE, SIZE)).astype(
+                        dtype[source]
+                    )
+                    f.write(data, j)
+
+            if source == "s2":
+                file_list.append(filename)
+
+        # Create tarballs
+        shutil.make_archive(f"{season}_{source}", "gztar", ".", directory)
+
+        # Compute checksums
+        with open(tarball, "rb") as f:
+            md5 = hashlib.md5(f.read()).hexdigest()
+            print(repr(md5) + ",")
+
+for split in ["train", "test"]:
+    filename = f"{split}_list.txt"
+
+    # Create file list
+    with open(filename, "w") as f:
+        for fname in file_list:
+            f.write(f"{fname}\n")
+
+    # Compute checksums
+    with open(filename, "rb") as f:
+        md5 = hashlib.md5(f.read()).hexdigest()
+        print(repr(md5) + ",")

tests/datasets/test_sen12ms.py

@@ -22,18 +22,20 @@ def dataset(
         self, monkeypatch: Generator[MonkeyPatch, None, None], request: SubRequest
     ) -> SEN12MS:
         md5s = [
-            "7f14be13d3f62c09b4dd5b4d55c97fd6",
-            "48182d44b375360381f36d432956b225",
-            "96cf1b8405d4149c6fe61ad7100bd65d",
-            "ba8e7e10fba9eea6900ddc530c86025a",
-            "7ba7c51f2fb3a2074b7bbd3e24f9d70d",
-            "280c9be2d1e13e663824dccd85e1e42f",
-            "a5284baf48534d4bc77acb1b103ff16c",
-            "c6b176fed0cdd5033cb1835506e40ee4",
-            "adc672746b79be4c4edc8b1a564e3ff4",
-            "194fab4a4e067a0452824c4e39f61b77",
-            "7899c0c36c884ae8c991ab8518b0d177",
-            "ccfee543d4351bcc5aa68729e8cc795c",
+            "b7d9e183a460979e997b443517a78ded",
+            "7131dbb098c832fff84c2b8a0c8f1126",
+            "b1057fea6ced6d648e5b16efeac352ad",
+            "2da32111fcfb80939aea7b18c2250fa8",
+            "c688ad6475660dbdbc36f66a1dd07da7",
+            "2ecd0dce2a21372513955c604b07e24f",
+            "dbc84c03edf77a68f789a6f7d2ea66a9",
+            "3e42a7dc4bb1ecd8c588930bf49b5c2b",
+            "c29053cb8cf5d75e333b1b51d37f62fe",
+            "5b6880526bc6da488154092741392042",
+            "d1b51c39b1013f2779fecf1f362f6c28",
+            "078def1e13ce4e88632d65f5c73a6259",
+            "02d5128ac1fc2bf8762091b4f319762d",
+            "02d5128ac1fc2bf8762091b4f319762d",
         ]
 
         monkeypatch.setattr(SEN12MS, "md5s", md5s)  # type: ignore[attr-defined]

tests/samplers/test_batch.py

@@ -9,7 +9,7 @@
 from rasterio.crs import CRS
 from torch.utils.data import DataLoader
 
-from torchgeo.datasets import BoundingBox, GeoDataset
+from torchgeo.datasets import BoundingBox, GeoDataset, stack_samples
 from torchgeo.samplers import BatchGeoSampler, RandomBatchGeoSampler
 
 
@@ -51,7 +51,9 @@ def test_len(self, sampler: CustomBatchGeoSampler) -> None:
     @pytest.mark.parametrize("num_workers", [0, 1, 2])
     def test_dataloader(self, sampler: CustomBatchGeoSampler, num_workers: int) -> None:
         ds = CustomGeoDataset()
-        dl = DataLoader(ds, batch_sampler=sampler, num_workers=num_workers)
+        dl = DataLoader(
+            ds, batch_sampler=sampler, num_workers=num_workers, collate_fn=stack_samples
+        )
         for _ in dl:
             continue
 
@@ -100,6 +102,8 @@ def test_roi(self) -> None:
     @pytest.mark.parametrize("num_workers", [0, 1, 2])
     def test_dataloader(self, sampler: RandomBatchGeoSampler, num_workers: int) -> None:
         ds = CustomGeoDataset()
-        dl = DataLoader(ds, batch_sampler=sampler, num_workers=num_workers)
+        dl = DataLoader(
+            ds, batch_sampler=sampler, num_workers=num_workers, collate_fn=stack_samples
+        )
         for _ in dl:
             continue

tests/samplers/test_single.py

@@ -9,7 +9,7 @@
 from rasterio.crs import CRS
 from torch.utils.data import DataLoader
 
-from torchgeo.datasets import BoundingBox, GeoDataset
+from torchgeo.datasets import BoundingBox, GeoDataset, stack_samples
 from torchgeo.samplers import GeoSampler, GridGeoSampler, RandomGeoSampler
 
 
@@ -55,7 +55,9 @@ def test_abstract(self) -> None:
     @pytest.mark.parametrize("num_workers", [0, 1, 2])
     def test_dataloader(self, sampler: CustomGeoSampler, num_workers: int) -> None:
         ds = CustomGeoDataset()
-        dl = DataLoader(ds, sampler=sampler, num_workers=num_workers)
+        dl = DataLoader(
+            ds, sampler=sampler, num_workers=num_workers, collate_fn=stack_samples
+        )
         for _ in dl:
             continue
 
@@ -97,7 +99,9 @@ def test_roi(self) -> None:
     @pytest.mark.parametrize("num_workers", [0, 1, 2])
     def test_dataloader(self, sampler: RandomGeoSampler, num_workers: int) -> None:
         ds = CustomGeoDataset()
-        dl = DataLoader(ds, sampler=sampler, num_workers=num_workers)
+        dl = DataLoader(
+            ds, sampler=sampler, num_workers=num_workers, collate_fn=stack_samples
+        )
         for _ in dl:
             continue
 
@@ -145,6 +149,8 @@ def test_roi(self) -> None:
     @pytest.mark.parametrize("num_workers", [0, 1, 2])
     def test_dataloader(self, sampler: GridGeoSampler, num_workers: int) -> None:
         ds = CustomGeoDataset()
-        dl = DataLoader(ds, sampler=sampler, num_workers=num_workers)
+        dl = DataLoader(
+            ds, sampler=sampler, num_workers=num_workers, collate_fn=stack_samples
+        )
         for _ in dl:
             continue

tests/trainers/test_byol.py

@@ -14,6 +14,8 @@
 from torchgeo.trainers import BYOLTask
 from torchgeo.trainers.byol import BYOL, SimCLRAugmentation
 
+from .test_utils import ClassificationTestModel
+
 
 class TestBYOL:
     def test_custom_augment_fn(self) -> None:
@@ -53,6 +55,8 @@ def test_trainer(self, name: str, classname: Type[LightningDataModule]) -> None:
         model_kwargs = conf_dict["module"]
         model = BYOLTask(**model_kwargs)
 
+        model.encoder = ClassificationTestModel(**model_kwargs)
+
         # Instantiate trainer
         trainer = Trainer(fast_dev_run=True, log_every_n_steps=1)
         trainer.fit(model=model, datamodule=datamodule)

tests/trainers/test_classification.py

@@ -2,11 +2,14 @@
 # Licensed under the MIT License.
 
 import os
-from typing import Any, Dict, Type, cast
+from typing import Any, Dict, Generator, Type, cast
 
 import pytest
+import timm
+from _pytest.monkeypatch import MonkeyPatch
 from omegaconf import OmegaConf
 from pytorch_lightning import LightningDataModule, Trainer
+from torch.nn.modules import Module
 
 from torchgeo.datamodules import (
     BigEarthNetDataModule,
@@ -17,6 +20,12 @@
 )
 from torchgeo.trainers import ClassificationTask, MultiLabelClassificationTask
 
+from .test_utils import ClassificationTestModel
+
+
+def create_model(*args: Any, **kwargs: Any) -> Module:
+    return ClassificationTestModel(**kwargs)
+
 
 class TestClassificationTask:
     @pytest.mark.parametrize(
@@ -29,8 +38,13 @@ class TestClassificationTask:
             ("ucmerced", UCMercedDataModule),
         ],
     )
-    def test_trainer(self, name: str, classname: Type[LightningDataModule]) -> None:
-        if name == "so2sat":
+    def test_trainer(
+        self,
+        monkeypatch: Generator[MonkeyPatch, None, None],
+        name: str,
+        classname: Type[LightningDataModule],
+    ) -> None:
+        if name.startswith("so2sat"):
             pytest.importorskip("h5py")
 
         conf = OmegaConf.load(os.path.join("conf", "task_defaults", name + ".yaml"))
@@ -42,6 +56,9 @@ def test_trainer(self, name: str, classname: Type[LightningDataModule]) -> None:
         datamodule = classname(**datamodule_kwargs)
 
         # Instantiate model
+        monkeypatch.setattr(  # type: ignore[attr-defined]
+            timm, "create_model", create_model
+        )
         model_kwargs = conf_dict["module"]
         model = ClassificationTask(**model_kwargs)
 
@@ -119,7 +136,12 @@ class TestMultiLabelClassificationTask:
             ("bigearthnet_s2", BigEarthNetDataModule),
         ],
     )
-    def test_trainer(self, name: str, classname: Type[LightningDataModule]) -> None:
+    def test_trainer(
+        self,
+        monkeypatch: Generator[MonkeyPatch, None, None],
+        name: str,
+        classname: Type[LightningDataModule],
+    ) -> None:
         conf = OmegaConf.load(os.path.join("conf", "task_defaults", name + ".yaml"))
         conf_dict = OmegaConf.to_object(conf.experiment)
         conf_dict = cast(Dict[Any, Dict[Any, Any]], conf_dict)
@@ -129,6 +151,9 @@ def test_trainer(self, name: str, classname: Type[LightningDataModule]) -> None:
         datamodule = classname(**datamodule_kwargs)
 
         # Instantiate model
+        monkeypatch.setattr(  # type: ignore[attr-defined]
+            timm, "create_model", create_model
+        )
         model_kwargs = conf_dict["module"]
         model = MultiLabelClassificationTask(**model_kwargs)
 

tests/trainers/test_regression.py

@@ -11,6 +11,8 @@
 from torchgeo.datamodules import COWCCountingDataModule, CycloneDataModule
 from torchgeo.trainers import RegressionTask
 
+from .test_utils import RegressionTestModel
+
 
 class TestRegressionTask:
     @pytest.mark.parametrize(
@@ -30,6 +32,8 @@ def test_trainer(self, name: str, classname: Type[LightningDataModule]) -> None:
         model_kwargs = conf_dict["module"]
         model = RegressionTask(**model_kwargs)
 
+        model.model = RegressionTestModel()
+
         # Instantiate trainer
         trainer = Trainer(fast_dev_run=True, log_every_n_steps=1)
         trainer.fit(model=model, datamodule=datamodule)