python3.11 compatibility; is_tll to mtd_kde (#39)

TY-Cheng · web-flow · commit 958f51ff6a36 · 2025-06-18T11:46:53.000+08:00
* python3.11 compatibility; is_tll to mtd_kde

modify fstring for python3.11 compatibility;
is_tll to mtd_kde for more KDE to come.
diff --git a/.github/workflows/publish-package.yml b/.github/workflows/publish-package.yml
@@ -17,10 +17,10 @@ jobs:
       - name: Checkout code
         uses: actions/checkout@v4
 
-      - name: Set up Python 3.12
+      - name: Set up Python 3.11
         uses: actions/setup-python@v5
         with:
-          python-version: 3.12
+          python-version: 3.11
 
       - name: Cache pip
         uses: actions/cache@v4
diff --git a/.github/workflows/python-package.yml b/.github/workflows/python-package.yml
@@ -18,7 +18,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest, windows-latest, macos-latest]
-        python-version: ["3.12", "3.13"]
+        python-version: ["3.11", "3.12", "3.13"]
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
diff --git a/.github/workflows/static.yml b/.github/workflows/static.yml
@@ -28,10 +28,10 @@ jobs:
       - name: Checkout code
         uses: actions/checkout@v4
 
-      - name: Set up Python 3.12
+      - name: Set up Python 3.11
         uses: actions/setup-python@v5
         with:
-          python-version: "3.12"
+          python-version: "3.11"
 
       - name: Cache pip
         uses: actions/cache@v4
diff --git a/README.md b/README.md
@@ -38,9 +38,9 @@ pip install torchvinecopulib torch
 
 ```bash
 # inside project root folder
-pip install ./dist/torchvinecopulib-1.0.4-py3-none-any.whl
+pip install ./dist/torchvinecopulib-1.1.0-py3-none-any.whl
 # or
-pip install ./dist/torchvinecopulib-1.0.4.tar.gz
+pip install ./dist/torchvinecopulib-1.1.0.tar.gz
 ```
 
 ### (Recommended) [uv](https://docs.astral.sh/uv/getting-started/) for Dependency Management and Packaging
@@ -61,7 +61,7 @@ uv sync --extra cu126 -U
 fastkde = "*"
 numpy = "*"
 pyvinecopulib = "*"
-python = ">=3.12"
+python = ">=3.11"
 scipy = "*"
 # optional to facilitate customization
 torch = [
diff --git a/examples/0_bicop.ipynb b/examples/0_bicop.ipynb
diff --git a/examples/1_vinecop.ipynb b/examples/1_vinecop.ipynb
diff --git a/examples/3_bicop_tll.ipynb b/examples/3_bicop_tll.ipynb
diff --git a/examples/4_mnist.ipynb b/examples/4_mnist.ipynb
@@ -530,7 +530,7 @@
     ").to(DEVICE)\n",
     "vine_tvc.fit(\n",
     "    obs=representations_subset,\n",
-    "    is_tll=True,\n",
+    "    mtd_kde='tll',\n",
     ")\n",
     "# sns.histplot(\n",
     "#     x=vine_tvc.log_pdf(representations_subset).cpu().numpy().reshape(-1),\n",
@@ -711,7 +711,7 @@
     "        obs=u_cls_tensor,\n",
     "        matrix=tvc_mat,\n",
     "        is_dissmann=False,\n",
-    "        is_tll=True,\n",
+    "        mtd_kde='tll',\n",
     "    )\n",
     "\n",
     "    # 4. Store the fitted model\n",
@@ -762,7 +762,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": null,
    "id": "c3b24468",
    "metadata": {},
    "outputs": [
@@ -781,14 +781,8 @@
     ")\n",
     "\n",
     "\n",
-    "bicop_tvc = tvc.BiCop(\n",
-    "    num_step_grid=30,\n",
-    ").to(DEVICE)\n",
-    "\n",
-    "bicop_tvc.fit(\n",
-    "    obs=u_cls_tensor[:, :2],\n",
-    "    is_tll=True,\n",
-    ")\n",
+    "bicop_tvc = tvc.BiCop(num_step_grid=30).to(DEVICE)\n",
+    "bicop_tvc.fit(obs=u_cls_tensor[:, :2], mtd_kde=\"tll\")\n",
     "print(bicop_pvc.loglik(), bicop_tvc.log_pdf(obs=u_cls_tensor[:, :2]).sum())"
    ]
   },
diff --git a/examples/pred_intvl/vine_wrapper.py b/examples/pred_intvl/vine_wrapper.py
@@ -16,7 +16,7 @@ def train_vine(Z_train, Y_train, seed=42, device="cpu"):
         obs=ZY_train,
         first_tree_vertex=first_tree_vertex,
         mtd_bidep="ferreira_tail_dep_coeff",
-        is_tll=True,
+        mtd_kde="tll",
         mtd_tll="quadratic",
         seed=seed,
     )
diff --git a/examples/vcae_mnist/vcae_mnist/model.py b/examples/vcae_mnist/vcae_mnist/model.py
@@ -210,7 +210,7 @@ def learn_vine(self, n_samples=5000):
         ).to(DEVICE)
         vine_tvc.fit(
             obs=representations_subset,
-            is_tll=True,
+            mtd_kde="tll",
         )
         self.set_vine(vine_tvc)
         self.use_vine = True
diff --git a/pyproject.toml b/pyproject.toml
@@ -17,8 +17,8 @@ documentation = "https://ty-cheng.github.io/torchvinecopulib/"
 
 [project]
 name = "torchvinecopulib"
-version = "1.0.4"
-requires-python = ">=3.12"
+version = "1.1.0"
+requires-python = ">=3.11"
 authors = [{ name = "Tuoyuan Cheng", email = "tuoyuan.cheng@nus.edu.sg" }]
 maintainers = [
     { name = "Saikiran Reddy Poreddy", email = "saikiran.poreddy@u.nus.edu" },
@@ -39,6 +39,7 @@ classifiers = [
     "Operating System :: Microsoft :: Windows :: Windows 10",
     "Operating System :: POSIX :: Linux",
     "Operating System :: MacOS",
+    "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
     "Programming Language :: Python :: 3.13",
     "Programming Language :: Python :: 3.14",
diff --git a/tests/__init__.py b/tests/__init__.py
@@ -40,10 +40,10 @@ def bicop_pair(request):
 
     # 2) fit two torchvinecopulib instances (fast KDE and TLL)
     bc_fast = tvc.BiCop(num_step_grid=512).to(DEVICE)
-    bc_fast.fit(U_tensor, is_tll=False)
+    bc_fast.fit(U_tensor, mtd_kde="fastKDE")
 
     bc_tll = tvc.BiCop(num_step_grid=512).to(DEVICE)
-    bc_tll.fit(U_tensor, is_tll=True)
+    bc_tll.fit(U_tensor, mtd_kde="tll")
 
     return family, true_params, rotation, U_tensor, bc_fast, bc_tll
 
diff --git a/tests/test_bicop.py b/tests/test_bicop.py
@@ -118,10 +118,10 @@ def test_sample_marginals(bicop_pair):
 
 def test_internal_buffers_and_flags(bicop_pair):
     _, _, _, U, bc_fast, bc_tll = bicop_pair
-    for cop, expect_tll in [(bc_fast, False), (bc_tll, True)]:
+    for cop, mtd_kde in [(bc_fast, "fastKDE"), (bc_tll, "tll")]:
         print(cop)
         assert not cop.is_indep
-        assert cop.is_tll is expect_tll
+        assert cop.mtd_kde == mtd_kde
         assert cop.num_obs == U.shape[0]
         # all the pre‐computed grids are the right shape
         m = cop.num_step_grid
@@ -131,10 +131,10 @@ def test_internal_buffers_and_flags(bicop_pair):
 
 
 def test_tau_estimation(bicop_pair):
-    _, _, _, U, bc_fast, bc_tll = bicop_pair
+    _, _, _, U, bc_fast, bc_mtd_kde = bicop_pair
     # re‐fit with tau estimation
     bc = tvc.BiCop(num_step_grid=64)
-    bc.fit(U, is_tll=True, is_tau_est=True)
+    bc.fit(U, mtd_kde="tll", is_tau_est=True)
     # kendalltau must be nonzero for dependent data
     assert bc.tau[0].abs().item() > 0
     assert bc.tau[1].abs().item() >= 0
@@ -156,20 +156,28 @@ def test_imshow_and_plot_api(bicop_pair):
     fig, ax = cop.imshow(is_log_pdf=True)
     assert isinstance(fig, matplotlib.figure.Figure)
     assert isinstance(ax, matplotlib.axes.Axes)
+    plt.close(fig)
+
     # contour
     fig2, ax2 = cop.plot(plot_type="contour", margin_type="unif")
     assert isinstance(fig2, matplotlib.figure.Figure)
     assert isinstance(ax2, matplotlib.axes.Axes)
+    plt.close(fig2)
     fig2, ax2 = cop.plot(plot_type="contour", margin_type="norm")
     assert isinstance(fig2, matplotlib.figure.Figure)
     assert isinstance(ax2, matplotlib.axes.Axes)
+    plt.close(fig2)
+
     # surface
     fig3, ax3 = cop.plot(plot_type="surface", margin_type="unif")
     assert isinstance(fig3, matplotlib.figure.Figure)
     assert isinstance(ax3, matplotlib.axes.Axes)
+    plt.close(fig3)
     fig3, ax3 = cop.plot(plot_type="surface", margin_type="norm")
     assert isinstance(fig3, matplotlib.figure.Figure)
     assert isinstance(ax3, matplotlib.axes.Axes)
+    plt.close(fig3)
+
     # invalid args
     with pytest.raises(ValueError):
         cop.plot(plot_type="foo")
@@ -194,21 +202,21 @@ def test_reset_and_str(bicop_pair):
         assert cop.num_obs == 0
         # __str__ contains key fields
         s = str(cop)
-        assert "is_indep" in s and "num_obs" in s and "is_tll" in s
+        assert "is_indep" in s and "num_obs" in s and "mtd_kde" in s
 
 
 @pytest.mark.parametrize("method", ["constant", "linear", "quadratic"])
 def test_tll_methods_do_not_crash(U_tensor, method):
     cop = tvc.BiCop(num_step_grid=32)
     # should _not_ raise for any of the valid nonparametric_method names
-    cop.fit(U_tensor, is_tll=True, mtd_tll=method)
+    cop.fit(U_tensor, mtd_kde="tll", mtd_tll=method)
 
 
 def test_fit_invalid_method_raises(U_tensor):
     cop = tvc.BiCop(num_step_grid=32)
     with pytest.raises(RuntimeError):
         # pick something bogus
-        cop.fit(U_tensor, is_tll=True, mtd_tll="no_such_method")
+        cop.fit(U_tensor, mtd_kde="tll", mtd_tll="no_such_method")
 
 
 def test_interp_on_trivial_grid():
@@ -253,11 +261,13 @@ def test_interp_on_trivial_grid():
 def test_imshow_with_existing_axes():
     cop = tvc.BiCop(num_step_grid=32)
     us = torch.rand(100, 2)
-    cop.fit(us, is_tll=False)
+    cop.fit(us, mtd_kde="fastKDE")
     fig, outer_ax = plt.subplots()
     fig2, ax2 = cop.imshow(is_log_pdf=False, ax=outer_ax, cmap="viridis")
     # should have returned the same axes object
     assert ax2 is outer_ax
+    plt.close(fig)
+    plt.close(fig2)
 
 
 def test_independent_copula_properties():
diff --git a/tests/test_vinecop.py b/tests/test_vinecop.py
@@ -1,3 +1,4 @@
+import matplotlib.pyplot as plt
 import networkx as nx
 import pytest
 import torch
@@ -46,7 +47,7 @@ def test_fit_sample_logpdf_cdf_forward(mtd_vine, mtd_bidep):
     # fit in copula-scale
     vc.fit(
         obs=U,
-        is_tll=True,  # use TLL for bidep estimation
+        mtd_kde="tll",  # use TLL for bidep estimation
         is_dissmann=True,
         mtd_vine=mtd_vine,
         mtd_bidep=mtd_bidep,
@@ -84,7 +85,7 @@ def test_matrix_diagonal_matches_sample_order(mtd_vine, mtd_bidep):
     U = torch.rand(100, num_dim, dtype=torch.float64)
     vc.fit(
         U,
-        is_tll=True,
+        mtd_kde="tll",
         is_dissmann=True,
         mtd_vine=mtd_vine,
         mtd_bidep=mtd_bidep,
@@ -107,7 +108,7 @@ def test_fit_with_explicit_matrix_uses_exact_edges():
     torch.manual_seed(1)
     U = torch.rand(200, num_dim, dtype=torch.float64)
     vc = VineCop(num_dim, is_cop_scale=True, num_step_grid=16)
-    vc.fit(U, is_tll=True, is_dissmann=True)
+    vc.fit(U, mtd_kde="tll", is_dissmann=True)
     M = vc.matrix
     # M is a square matrix with num_dim rows and columns
     vc = VineCop(num_dim, is_cop_scale=True, num_step_grid=16)
@@ -144,7 +145,7 @@ def test_reset_clears_all_levels_and_bicops():
     vc.fit(
         U,
         is_dissmann=True,
-        is_tll=True,
+        mtd_kde="tll",
         mtd_vine="rvine",
         mtd_bidep="kendall_tau",
         thresh_trunc=None,
@@ -167,7 +168,7 @@ def test_reset_and_str():
     vc.fit(
         U,
         is_dissmann=True,
-        is_tll=True,
+        mtd_kde="tll",
         mtd_vine="rvine",
         mtd_bidep="kendall_tau",
         thresh_trunc=None,
@@ -200,7 +201,7 @@ def test_ref_count_hfunc_on_fitted_vine(mtd_vine, mtd_bidep):
     vc = VineCop(num_dim, is_cop_scale=True, num_step_grid=16)
     vc.fit(
         U,
-        is_tll=True,
+        mtd_kde="tll",
         is_dissmann=True,
         mtd_vine=mtd_vine,
         mtd_bidep=mtd_bidep,
@@ -229,7 +230,7 @@ def test_draw_lv_and_draw_dag(tmp_path):
     vc = VineCop(num_dim, is_cop_scale=True, num_step_grid=32)
     vc.fit(
         U,
-        is_tll=True,
+        mtd_kde="tll",
         is_dissmann=True,
         mtd_vine="rvine",
         mtd_bidep="kendall_tau",
@@ -240,25 +241,30 @@ def test_draw_lv_and_draw_dag(tmp_path):
     fig, ax, G = vc.draw_lv(lv=0, is_bcp=False)
     assert isinstance(G, nx.Graph)
     assert G.number_of_nodes() > 0
+    plt.close(fig)
 
     # save level-1 bicop view
     fpath_lv = tmp_path / "level1.png"
     fig2, ax2, G2, outpath_lv = vc.draw_lv(lv=1, is_bcp=True, f_path=fpath_lv)
     assert outpath_lv == fpath_lv
     assert fpath_lv.exists()
+    plt.close(fig2)
     # save level-1 pseudo-obs view
     fpath_lv = tmp_path / "level1.png"
     fig2, ax2, G2, outpath_lv = vc.draw_lv(lv=1, is_bcp=False, f_path=fpath_lv)
     assert outpath_lv == fpath_lv
     assert fpath_lv.exists()
+    plt.close(fig2)
 
     # draw the DAG
     fig3, ax3, G3 = vc.draw_dag()
     assert isinstance(G3, nx.DiGraph)
     assert len(G3.nodes) > 0
+    plt.close(fig3)
 
     # save DAG
     fpath_dag = tmp_path / "dag.png"
     fig4, ax4, G4, outpath_dag = vc.draw_dag(f_path=fpath_dag)
     assert outpath_dag == fpath_dag
     assert fpath_dag.exists()
+    plt.close(fig4)
diff --git a/torchvinecopulib/bicop/__init__.py b/torchvinecopulib/bicop/__init__.py
diff --git a/torchvinecopulib/vinecop/__init__.py b/torchvinecopulib/vinecop/__init__.py

Original file line number	Diff line number	Diff line change
`@@ -16,7 +16,7 @@ def train_vine(Z_train, Y_train, seed=42, device="cpu"):`
`16`	`16`	`obs=ZY_train,`
`17`	`17`	`first_tree_vertex=first_tree_vertex,`
`18`	`18`	`mtd_bidep="ferreira_tail_dep_coeff",`
`19`		`- is_tll=True,`
	`19`	`+ mtd_kde="tll",`
`20`	`20`	`mtd_tll="quadratic",`
`21`	`21`	`seed=seed,`
`22`	`22`	`)`
Original file line number	Diff line number	Diff line change
`@@ -210,7 +210,7 @@ def learn_vine(self, n_samples=5000):`
`210`	`210`	`).to(DEVICE)`
`211`	`211`	`vine_tvc.fit(`
`212`	`212`	`obs=representations_subset,`
`213`		`- is_tll=True,`
	`213`	`+ mtd_kde="tll",`
`214`	`214`	`)`
`215`	`215`	`self.set_vine(vine_tvc)`
`216`	`216`	`self.use_vine = True`