[pre-commit.ci] pre-commit autoupdate

.pre-commit-config.yaml

@@ -7,14 +7,14 @@ default_install_hook_types: [pre-commit, commit-msg]
 
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.1.9
+    rev: v0.3.4
     hooks:
       - id: ruff
         args: [--fix]
       - id: ruff-format
 
   - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.8.0
+    rev: v1.9.0
     hooks:
       - id: mypy
 
@@ -39,7 +39,7 @@ repos:
         args: [--check-filenames]
 
   - repo: https://github.com/kynan/nbstripout
-    rev: 0.6.1
+    rev: 0.7.1
     hooks:
       - id: nbstripout
         args: [--drop-empty-cells, --keep-output]

examples/functorch_mlp_ensemble.ipynb

@@ -164,7 +164,8 @@
     "\n",
     "    plt.scatter(*points.T, c=labels)\n",
     "    plt.gca().set(xlabel=\"x\", ylabel=\"y\", title=f\"{noise_std=}\")\n",
-    "    plt.show()"
+    "    plt.show()\n",
+    "\n"
    ]
   },
   {
@@ -297,6 +298,8 @@
    ],
    "source": [
     "# If so, the loss should decrease with step count.\n",
+    "\n",
+    "\n",
     "metrics = {}\n",
     "for step in range(n_train_steps):\n",
     "    loss, acc, weights = train_step_fn(weights, points, labels)\n",
@@ -310,7 +313,8 @@
     "ax.set_ylabel(\"loss\", color=\"tab:blue\")\n",
     "ax2 = ax.twinx()\n",
     "ax2.plot(list(metrics), [v[\"acc\"] for v in metrics.values()], color=\"red\")\n",
-    "ax2.set_ylabel(\"accuracy\", color=\"tab:red\")"
+    "ax2.set_ylabel(\"accuracy\", color=\"tab:red\")\n",
+    "\n"
    ]
   },
   {
@@ -352,6 +356,8 @@
    "source": [
     "# at the same time! Note that metrics like losses and accuracies are all tuples here,\n",
     "# one scalar per model.\n",
+    "\n",
+    "\n",
     "parallel_train_step_fn = functorch.vmap(train_step_fn, in_dims=(0, None, None))\n",
     "batched_weights = initialize_ensemble(n_models=5)\n",
     "for step in tqdm(range(n_train_steps), desc=\"training MLP ensemble\"):\n",

examples/wandb_integration.ipynb

@@ -183,7 +183,8 @@
     "    \"learning_rate\": 0.005,\n",
     "    \"dataset\": \"MNIST\",\n",
     "    \"architecture\": \"CNN\",\n",
-    "}"
+    "}\n",
+    "\n"
    ]
   },
   {
@@ -281,7 +282,8 @@
     "        out = self.layer1(x)\n",
     "        out = self.layer2(out)\n",
     "        out = out.reshape(out.size(0), -1)\n",
-    "        return self.fc(out)"
+    "        return self.fc(out)\n",
+    "\n"
    ]
   },
   {
@@ -371,7 +373,8 @@
     "                wandb.log(\n",
     "                    {\"epoch\": epoch, \"loss\": loss, \"accuracy\": accuracy},\n",
     "                    step=sample_count,\n",
-    "                )"
+    "                )\n",
+    "\n"
    ]
   },
   {
@@ -474,6 +477,8 @@
    "outputs": [],
    "source": [
     "# Make the data\n",
+    "\n",
+    "\n",
     "train_set, test_set = (\n",
     "    torchvision.datasets.MNIST(\n",
     "        root=\".\",\n",
@@ -515,7 +520,8 @@
     "\n",
     "    wandb.finish()\n",
     "\n",
-    "    return model"
+    "    return model\n",
+    "\n"
    ]
   },
   {

paper/references.bib

@@ -11,7 +11,7 @@ @online{biewald_weights_2020
   date = {2020},
   url = {https://docs.wandb.ai/company/academics},
   urldate = {2022-08-21},
-  organization = {{Use Weights \& Biases for free to track experiments, collaborate, and publish results}}
+  organization = {Use Weights \& Biases for free to track experiments, collaborate, and publish results}
 }
 
 @software{bradbury_jax_2018,
@@ -29,7 +29,7 @@ @software{developers_tensorflow_2022
   doi = {10.5281/zenodo.6574269},
   url = {https://zenodo.org/record/6574269},
   urldate = {2022-08-21},
-  organization = {{Zenodo}}
+  organization = {Zenodo}
 }
 
 @online{fey_fast_2019,
@@ -55,7 +55,7 @@ @article{harris_array_2020
   volume = {585},
   number = {7825},
   pages = {357--362},
-  publisher = {{Nature Publishing Group}},
+  publisher = {Nature Publishing Group},
   issn = {1476-4687},
   doi = {10.1038/s41586-020-2649-2},
   url = {https://www.nature.com/articles/s41586-020-2649-2},
@@ -72,7 +72,7 @@ @article{jumper_highly_2021
   volume = {596},
   number = {7873},
   pages = {583--589},
-  publisher = {{Nature Publishing Group}},
+  publisher = {Nature Publishing Group},
   issn = {1476-4687},
   doi = {10.1038/s41586-021-03819-2},
   url = {https://www.nature.com/articles/s41586-021-03819-2},

pyproject.toml

@@ -65,9 +65,11 @@ warn_unused_ignores = true
 [tool.ruff]
 target-version = "py38"
 include = ["**/pyproject.toml", "*.ipynb", "*.py", "*.pyi"]
+
+[tool.ruff.lint]
 select = ["ALL"]
 ignore = [
-  "ANN101",
+  "ANN101", # Missing type annotation for self in method
   "ANN401",
   "ARG001",
   "C901",
@@ -87,7 +89,7 @@ ignore = [
 ]
 pydocstyle.convention = "google"
 
-[tool.ruff.per-file-ignores]
+[tool.ruff.lint.per-file-ignores]
 "tests/*" = ["D103", "D104", "INP001", "S101"]
 "__init__.py" = ["F401"]
 "examples/*" = ["D102", "D103", "D107", "E402", "FA102"]

tensorboard_reducer/write.py

@@ -33,7 +33,7 @@ def _rm_rf_or_raise(path: str, overwrite: bool) -> None:
         if overwrite and (is_data_file or is_tb_dir):
             os.system(f"rm -rf {path}")  # noqa: S605
         elif overwrite:
-            ValueError(
+            raise ValueError(
                 f"Received the overwrite flag but the content of '{path}' does not "
                 "look like it was written by this program. Please make sure you really "
                 f"want to delete '{path}' and then do so manually."

tests/test_reduce.py

@@ -13,8 +13,9 @@ def generate_sample_data(
     n_tags: int = 1, n_runs: int = 10, n_steps: int = 5
 ) -> dict[str, pd.DataFrame]:
     events_dict = {}
+    rng = np.random.default_rng()
     for idx in range(n_tags):
-        data = np.random.random((n_steps, n_runs))
+        data = rng.random((n_steps, n_runs))
         df_rand = pd.DataFrame(data, columns=[f"run_{j}" for j in range(n_runs)])
         events_dict[f"tag_{idx}"] = df_rand
     return events_dict
@@ -92,4 +93,4 @@ def test_reduce_events_dimensions(n_tags: int, n_runs: int, n_steps: int) -> Non
 @pytest.mark.parametrize("reduce_ops", [["mean"], ["max", "min"], ["std", "median"]])
 def test_reduce_events_empty_input(reduce_ops: Sequence[str]) -> None:
     reduced_events = reduce_events({}, reduce_ops)
-    assert reduced_events == dict.fromkeys(reduce_ops, {})
+    assert reduced_events == {op: {} for op in reduce_ops}