Merge pull request #1 from onnela-lab/reorg

Reorganize repository to use `uv`.

Author: tillahoffmann

.coveragerc

@@ -0,0 +1,5 @@
+[report]
+exclude_lines =
+    if __name__ == "__main__":
+    pragma: no cover
+    raise NotImplementedError

.github/workflows/main.yaml

@@ -1,43 +1,43 @@
-name: Summary Statistics
+name: CI
+
 on:
   push:
-    branches: [main]
+    branches: ["main"]
+    tags: ["*"]
   pull_request:
-    branches: [main]
+    branches: ["main"]
+  workflow_dispatch:
 
 env:
-  cmdstanVersion: "2.31.0"
+  cmdstanVersion: "2.36.0"
 
 jobs:
   build:
-    runs-on: ubuntu-latest
+    name: Continuous Integration
+    runs-on: "ubuntu-latest"
     steps:
-      - name: Check out the repository
-        uses: actions/checkout@v3
-      - name: Setup python
-        uses: actions/setup-python@v4
+      - uses: "actions/checkout@v4"
+      - name: Install uv
+        uses: astral-sh/setup-uv@v5
+        with:
+          enable-cache: true
+          cache-dependency-glob: uv.lock
+      - name: Set up Python
+        uses: actions/setup-python@v5
         with:
-          python-version: '3.10'
-          cache: pip
-      - name: Pre-install packages that are build requirements.
-        run: |
-          pip install `grep pybind11== requirements.txt` `grep wheel== requirements.txt`
-          pip install --no-build-isolation `grep -oE 'git\+.*?fasttr.git@\w+' requirements.txt`
-          pip install --extra-index-url https://download.pytorch.org/whl/cpu --no-compile --no-deps \
-            `grep torch== requirements.txt` \
-            `grep typing-extensions== requirements.txt` \
-            `grep mpmath== requirements.txt` \
-            `grep sympy== requirements.txt`
-      - name: Install Python dependencies.
-        run: pip install --extra-index-url https://download.pytorch.org/whl/cpu --no-compile --no-deps -r requirements.txt
-      - name: Cache cmdstan.
-        uses: actions/cache@v3
+          python-version-file: .python-version
+      - name: Install project
+        run: uv sync --all-groups
+      - name: Cache cmdstan
+        uses: actions/cache@v4
         with:
           path: /home/runner/.cmdstan
           key: cmdstan-${{ env.cmdstanVersion }}
-      - name: Install cmdstan.
-        run: python -m cmdstanpy.install_cmdstan --version ${{ env.cmdstanVersion }}
-      - name: Lint the code.
-        run: flake8
-      - name: Run the tests.
-        run: pytest --cov=summaries --cov-report=term-missing --cov-fail-under=100
+      - name: Install cmdstan
+        run: uv run python -m cmdstanpy.install_cmdstan --version ${{ env.cmdstanVersion }}
+      - name: Lint the code
+        run: true
+      - name: Run test_infer_mdn_tree separately
+        run: uv run pytest tests/scripts/test_infer_mdn_tree.py --cov=summaries --cov-report=term-missing
+      - name: Run all other tests
+        run: uv run pytest --cov=summaries --cov-report=term-missing --cov-fail-under=100 --cov-append --ignore=tests/scripts/test_infer_mdn_tree.py

.gitignore

@@ -163,5 +163,5 @@ workspace/
 *.ipynb
 *.tmp
 *.hpp
-summaries/scripts/infer_benchmark
+src/summaries/scripts/infer_benchmark
 ????-??-??-*

.python-version

@@ -0,0 +1 @@
+3.12

notebooks/benchmark.md

@@ -11,6 +11,7 @@ import torch
 
 
 mpl.style.use("../.mplstyle")
+figwidth, figheight = mpl.rcParams["figure.figsize"]
 ```
 
 ```python
@@ -67,7 +68,7 @@ X = torch.as_tensor(observed["data"][idx], dtype=dtype)
 x = X[:, 0]
 samples = mdnabc["samples"][idx].squeeze()
 
-fig, axes = plt.subplots(2, 2, sharex="col", sharey="row")
+fig, axes = plt.subplots(2, 2, sharex="col", sharey="row", figsize=(figwidth, figheight))
 
 ax = axes[0, 0] # ----- ----- ----- -----
 

notebooks/experiments.md

@@ -13,13 +13,13 @@ figwidth, figheight = mpl.rcParams["figure.figsize"]
 ```
 
 ```python
-fig, ax = plt.subplots()
+fig, ax = plt.subplots(figsize=(figwidth, figheight))
 
 vspace = 1
 hspace = 2.1
 kwargs = {
     "ha": "center",
-    "va": "center", 
+    "va": "center",
     "fontsize": "small",
 }
 fkwargs = kwargs | {
@@ -64,7 +64,7 @@ texts = {
     "mdn-compressed-samples": (2 * hspace, -7 * vspace, "MDN-compressed samples\n$\\tilde\\theta\\in\\mathbb{R}^p\\sim \\tilde f\\left(\\theta\\mid t(y)\\right)$", vkwargs),
     "estimator": (0, -6 * vspace, "mixture density\nnetwork $h:\\mathbb{R}^q\\rightarrow\\mathcal{F}$", fkwargs),
     "estimate": (0, -7 * vspace, "density estimate\n$\\hat f\\left(\\theta\\mid t(z)\\right)\\in \\mathcal{F}$", vkwargs),
-    "loss": (- hspace / 2, -8 * vspace, "NLP loss", fkwargs), 
+    "loss": (- hspace / 2, -8 * vspace, "NLP loss", fkwargs),
 }
 
 elements = {}
@@ -80,12 +80,12 @@ if True:
     ax.set_xlim(xmin, xmax)
     ax.set_ylim(ymin, ymax)
     ax.set_aspect("equal")
-    
+
     # Then adjust based on the actual extent of the box containing the text.
     fig.tight_layout()
     fig.draw_without_rendering()
     transform = ax.transData.inverted()
-    extents = np.asarray([transform.transform(element.get_bbox_patch().get_window_extent()) 
+    extents = np.asarray([transform.transform(element.get_bbox_patch().get_window_extent())
                           for element in elements.values()])
     xmin, ymin = extents.min(axis=0)[0]
     xmax, ymax = extents.max(axis=0)[1]
@@ -105,12 +105,12 @@ connections = [
     [("params", 6), ("simulator", 12)],
     [("simulator", 6), ("simulated_data", 12)],
     [
-        ("simulated_data", 3), 
+        ("simulated_data", 3),
         (get_anchor(elements["compressor"], 11).x, get_anchor(elements["simulated_data"], 3).y),
         ("compressor", 11),
     ],
     [
-        ("compressor", 7), 
+        ("compressor", 7),
         (get_anchor(elements["compressor"], 7).x, get_anchor(elements["simulated_summaries"], 2.75).y),
         ("simulated_summaries", 2.75),
     ],
@@ -119,7 +119,7 @@ connections = [
         (get_anchor(elements["compressor"], 6).x, get_anchor(elements["simulated_summaries"], 3.25).y),
         (get_anchor(elements["compressor"], 6).x, get_anchor(elements["abc"], 9).y),
         ("abc", 9),
-        
+
     ],
     [("simulated_summaries", 6), ("estimator", 12)],
     [("estimator", 6), ("estimate", 12)],
@@ -136,12 +136,12 @@ connections = [
 
     # Observed.
     [
-        ("observed_data", 9), 
+        ("observed_data", 9),
         (get_anchor(elements["compressor"], 1).x, get_anchor(elements["observed_data"], 9).y),
         ("compressor", 1),
     ],
     [
-        ("compressor", 5), 
+        ("compressor", 5),
         (get_anchor(elements["compressor"], 5).x, get_anchor(elements["observed_summaries"], 9).y),
         ("observed_summaries", 9),
     ],
@@ -198,7 +198,7 @@ for (row, col), cell in celld.items():
     x = cell.get_x()
     y = cell.get_y() - (row + 1) * (height - original_height)
     phantom = mpl.patches.Rectangle(
-        (x, y), cell.get_width(), cell.get_height(), transform=ax.transAxes, 
+        (x, y), cell.get_width(), cell.get_height(), transform=ax.transAxes,
         facecolor=color, zorder=-1)
     ax.add_patch(phantom)
 

notebooks/illustrations.md

@@ -15,7 +15,7 @@ figwidth, figheight = mpl.rcParams["figure.figsize"]
 # Illustration of different methods and classes of summaries
 
 ```python
-fig, axes = plt.subplots(1, 2, figsize=(figwidth, 3), width_ratios=(3, 2))
+fig, axes = plt.subplots(1, 2, figsize=(figwidth, 2.55), width_ratios=(3.2, 2))
 
 ax = axes[1]
 sets = {
@@ -126,7 +126,7 @@ for key, value in sets.items():
     ax.text(*value["textxy"], value["text"], **textkwargs, zorder=9)
 
 ax.set_xlim(0, 11)
-ax.set_ylim(-0.2, 10.6)
+ax.set_ylim(-3, 10.6)
 ax.set_aspect("equal")
 ax.set_axis_off()
 
@@ -157,10 +157,10 @@ elements = {}
 for key, (x, y, text) in texts.items():
     element = ax.text(x, y, text, ha="center", va="center", fontsize="small", bbox=bbox)
     elements[key] = element
-    
+
 hpad = 1.2
 ax.set_xlim(-hpad, hspace + hpad)
-ax.set_ylim(-2.5, 3.1)
+ax.set_ylim(-2.5, 2.5)
 ax.set_aspect("equal")
 
 ax.set_axis_off()
@@ -170,7 +170,7 @@ fig.draw_without_rendering()
 scale = ax.transAxes.get_matrix()[0, 0] / ax.transData.get_matrix()[0, 0]
 width = 2
 info = mpl.patches.FancyBboxPatch(
-    (-width / 2, -3.25 * info_vspace), width, 5.75 * info_vspace, boxstyle=boxstyle, mutation_scale=1 / scale, 
+    (-width / 2, -3.25 * info_vspace), width, 5.75 * info_vspace, boxstyle=boxstyle, mutation_scale=1 / scale,
     color="C0", alpha=0.2,
 )
 ax.add_patch(info)
@@ -181,8 +181,8 @@ fig.draw_without_rendering()
 info_right = get_anchor(info, 3).x
 paths = [
     ("--", [
-        get_anchor(info, 0), 
-        (get_anchor(info, 0).x, get_anchor(elements["approx"], 9).y), 
+        get_anchor(info, 0),
+        (get_anchor(info, 0).x, get_anchor(elements["approx"], 9).y),
         get_anchor(elements["approx"], 9)
     ]),
     ("--", [
@@ -212,7 +212,7 @@ for ls, path in paths:
     ax.add_patch(patch)
     arrow = mpl.patches.PathPatch(arrow_path(path, 0.075), color="gray")
     ax.add_patch(arrow)
-    
+
     # Add the bidirectional arrow for large-sample correspondence.
     if ls == "-":
         arrow = mpl.patches.PathPatch(arrow_path(path, 0.075, backward=True), color="gray")
@@ -227,7 +227,7 @@ labels = [
     "special case",
     "large-sample limit",
 ]
-ax.legend(handles, labels, loc="upper center")
+ax.legend(handles, labels, loc="lower left", bbox_to_anchor=(.9, 0))
 
 label_axes(axes)
 fig.savefig("../workspace/figures/illustration.pdf")

notebooks/prior-dependence.md

@@ -7,6 +7,7 @@ from summaries.entropy import estimate_mutual_information
 from scipy import stats
 
 mpl.style.use("../.mplstyle")
+figwidth, figheight = mpl.rcParams["figure.figsize"]
 ```
 
 ```python
@@ -55,7 +56,7 @@ num_points = 200  # Number of points in the figure (we sample more for MI estima
 
 results = generate_data(m, n, scale)
 
-fig, axes = plt.subplots(2, 2, sharex=True)
+fig, axes = plt.subplots(2, 2, sharex=True, figsize=(figwidth, 0.8 * figheight))
 
 # Show the two priors.
 ax = axes[1, 0]
@@ -73,7 +74,7 @@ ax = axes[0, 0]
 lin = np.linspace(-1, 1, 100) * (1 + 3 * scale)
 ax.plot(lin, np.maximum(0, lin), label=r'location', color='k')
 ax.plot(lin, np.minimum(np.exp(lin / 2), 1), label=r'scale', color='k', ls='--')
-ax.set_ylabel('likelihood parameters')
+ax.set_ylabel('likelihood\nparameters')
 
 # Plot the scatter of summaries against parameter value for both priors.
 step = m // num_points  # Only plot `num_points` for better visualisation.

notebooks/statistics.md

@@ -9,6 +9,7 @@ from pathlib import Path
 from scipy import stats
 
 mpl.style.use("../.mplstyle")
+figwidth, figheight = mpl.rcParams["figure.figsize"]
 ```
 
 ```python
@@ -173,7 +174,7 @@ pass
 ```
 
 ```python
-fig = plt.figure()
+fig = plt.figure(figsize=(figwidth, 0.9 * figheight))
 gs = fig.add_gridspec(2, 2)
 
 # Declare how each point should be visualized.

notebooks/tough-nuts.md

@@ -8,6 +8,7 @@ from snippets.plot import label_axes
 
 
 mpl.style.use("../.mplstyle")
+figwidth, figheight = mpl.rcParams["figure.figsize"]
 ```
 
 ```python
@@ -57,7 +58,7 @@ second_moment = np.linspace(.2, .725, 101)
 aa, ss = np.meshgrid(a, second_moment)
 entropy_gain = evaluate_entropy_gain(aa, b, n, ss)
 
-fig, axes = plt.subplots(1, 2)
+fig, axes = plt.subplots(1, 2, figsize=(figwidth, figheight * 0.8))
 
 # Plot entropy gain with "centered" colorbar.
 ax = axes[0]