v0.3.0: Rebuild & Pass tests

Author: xiachenrui

.github/workflows/build.yml

@@ -14,7 +14,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: [ 3.11 ]
+        python-version: [ "3.10", "3.11", "3.12", "3.13" ]
 
     steps:
       - name: Checkout repository
@@ -30,8 +30,9 @@ jobs:
           sudo apt update
           sudo apt install -y libxml2-utils pandoc
           curl -sSL https://install.python-poetry.org | python3 -
-          pip install --upgrade pip
-          pip install -e ".[docs, dev]"
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+          uv pip install -e ".[docs, dev]" --system
           install_pyg_dependencies
 
       - name: Build documentation

.github/workflows/release.yml

@@ -26,8 +26,9 @@ jobs:
           sudo apt update
           sudo apt install -y libxml2-utils pandoc
           curl -sSL https://install.python-poetry.org | python3 -
-          pip install --upgrade pip
-          pip install -e ".[docs, dev]"
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          echo "$HOME/.cargo/bin" >> $GITHUB_PATH
+          uv pip install -e ".[docs, dev]" --system
           install_pyg_dependencies
 
       - name: Build documentation

Dockerfile

@@ -8,7 +8,6 @@ RUN apt-get update \
     && rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
 
 # Copy DECIPHER package and install
-# USER rapids
 WORKDIR /app
 COPY ./decipher ./decipher
 COPY ./README.md ./README.md

decipher/cls.py

@@ -93,6 +93,9 @@ def register_data(
             self.edge_index = edge_index
         np.save(self.work_dir / "edge_index.npy", self.edge_index.numpy())
 
+        # Save hyperparameters
+        OmegaConf.save(CFG, self.work_dir / "hyperparams.yaml")
+
     def fit_sc(self) -> None:
         r"""
         Fit on single cell data
@@ -133,7 +136,6 @@ def fit_omics(self) -> None:
         # save embeddings
         np.save(self.work_dir / "center_emb.npy", self.center_emb)
         np.save(self.work_dir / "nbr_emb.npy", self.nbr_emb)
-        OmegaConf.save(CFG, self.work_dir / "hyperparams.yaml")
         logger.info(f"Results saved to {self.work_dir}")
 
     def load(self, from_dir: str) -> None:
@@ -170,6 +172,7 @@ def fit_ddp(self, gpus: int = -1) -> None:
         max_gpus = torch.cuda.device_count()
         assert max_gpus > 1, "DDP requires at least 2 GPUs."
         gpus = min(gpus, max_gpus) if gpus > 0 else max_gpus
+        logger.info(f"Using {gpus} GPUs for DDP training.")
 
         # DDP fit omics
         subprocess.run(

decipher/explain/gene/mixin.py

@@ -7,8 +7,8 @@
 import ray
 import scanpy as sc
 import torch
-import yaml
 from loguru import logger
+from omegaconf import OmegaConf
 from scanpy._utils import check_nonnegative_integers
 from scipy.sparse import issparse
 from torch_geometric.data import Batch, Data
@@ -82,8 +82,7 @@ def train_gene_select(
         cfg.center_dim = self.center_emb.shape[1]
         cfg.nbr_dim = self.nbr_emb.shape[1]
         cfg.work_dir = str(work_dir)
-        with open(work_dir / "gene_select_config.yaml", "w") as f:
-            yaml.dump(cfg, f)
+        OmegaConf.save(cfg, work_dir / "gene_select_config.yaml")
         logger.debug(f"Gene select config: {cfg}")
 
         # prepare data

decipher/explain/regress/regression.py

@@ -6,6 +6,7 @@
 import pandas as pd
 import ray
 import torch
+from loguru import logger
 from omegaconf import OmegaConf
 from pytorch_lightning import LightningDataModule, LightningModule
 from rui_utils.torch.trainer import fit_and_inference
@@ -126,7 +127,11 @@ def calc_metric(self, pred_y, y):
             "test_r2_score": float(r2),
             "test_mae_score": float(mae),
         }
-        self.log_dict(self.test_metric, prog_bar=True)
+        try:
+            self.log_dict(self.test_metric, prog_bar=True)
+        except Exception as e:  # noqa
+            logger.error(f"Error logging test metrics: {e}")
+            print(self.test_metric)
 
 
 def train_regress(
@@ -173,6 +178,8 @@ def train_regress(
     config.trainer.model_dir = str(Path(config.work_dir) / save_dir)
     regress_model = DeepRegression(config, test_cell_type)
     fit_and_inference(regress_model, data, config.trainer)
+    if not hasattr(regress_model, "test_metric"):
+        regress_model.on_test_epoch_end()
     test_metric = regress_model.test_metric
 
     metric_path = Path(config.work_dir) / save_dir / "test_metric.json"

pyproject.toml

@@ -6,8 +6,8 @@ build-backend = "poetry.core.masonry.api"
 [tool.poetry]
 name = "cell-decipher"
 packages = [{include = "decipher"}]
-version = "0.2.2"
-description = "Spatial-omics data embedding and analysis"
+version = "0.3.0"
+description = "DECIPHER for learning disentangled cellular embeddings in large-scale heterogeneous spatial omics data"
 readme = "README.md"
 license = "MIT"
 authors = ["Chen-Rui Xia <[email protected]>"]
@@ -25,6 +25,7 @@ classifiers = [
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
     "Topic :: Scientific/Engineering :: Bio-Informatics"
 ]
 
@@ -43,13 +44,14 @@ bs4 = "*"
 rui_utils = "*"
 harmony-pytorch = "*"
 
-torch = ">=2.0"
+torch = ">=2.0, <2.9"
 torchvision = "*"
 torchaudio = "*"
 pytorch-lightning = ">=2.1"
 torch_geometric = ">=2.3.0"
 tensorboard = "*"
 einops = "*"
+transformers = "*"
 
 sphinx = {version = "<=6.2.1", optional = true}
 sphinx-autodoc-typehints = {version = "*", optional = true}
@@ -69,24 +71,18 @@ papermill = {version = "*", optional = true}
 jupyter = {version = "*", optional = true}
 jupyter_contrib_nbextensions = {version = "*", optional = true}
 jupytext = {version = "*", optional = true}
-ipykernel = {version = "*", optional = true}
 nbformat = {version = "*", optional = true}
 pre-commit = {version = "*", optional = true}
 black = {version = "*", optional = true}
 isort = {version = "*", optional = true}
 py-spy = {version = "*", optional = true}
-bokeh = {version = "*", optional = true}
-datatable = {version = "*", optional = true}
-parse = {version = "*", optional = true}
-plotly = {version = "*", optional = true}
-kaleido = {version = "*", optional = true}
 
 [tool.poetry.extras]
 docs = ["sphinx", "sphinx-autodoc-typehints", "sphinx-copybutton", "sphinx-intl",
         "nbsphinx", "sphinx-rtd-theme", "sphinx_gallery", "jinja2", "myst-parser"]
 dev = ["pytest", "pytest-sugar", "pytest-cov", "pytest-xdist",
-        "papermill", "jupyter","jupyter_contrib_nbextensions", "ipykernel", "nbformat", "jupytext",
-        "pre-commit", "black", "isort", "py-spy", "bokeh", "datatable", "parse", "plotly", "kaleido"]
+        "papermill", "jupyter","jupyter_contrib_nbextensions", "nbformat", "jupytext",
+        "pre-commit", "black", "isort", "py-spy"]
 
 [tool.pyright]
 include = ["decipher"]
@@ -106,8 +102,8 @@ source = ["decipher"]
 show_missing = true
 
 [tool.poetry.scripts]
-decipher_ddp_sc = "decipher.ddp:decipher_ddp_sc"
-decipher_ddp_spatial = "decipher.ddp:decipher_ddp_spatial"
+decipher_ddp_sc = "decipher.utils:decipher_ddp_sc"
+decipher_ddp_spatial = "decipher.utils:decipher_ddp_spatial"
 install_pyg_dependencies = "decipher.utils:install_pyg_dep"
 
 [tool.jupytext]

tests/test_explain.py

@@ -1,17 +1,16 @@
 import torch
 from torch_geometric.data import Batch, Data
 
+from decipher import CFG
 from decipher.explain.gene.gene_selection import train_GAE
 from decipher.explain.regress.mixin import train_regress
-from decipher.utils import GENESELECT_CFG, REGRESS_CFG
 
 
 def test_GAE_mimic():
-    GENESELECT_CFG.center_dim = 8
-    GENESELECT_CFG.expr_dim = 100
-    GENESELECT_CFG.work_dir = "./results/explain"
-    GENESELECT_CFG.gae_epochs = 2
-    GENESELECT_CFG.fit.epochs = 2
+    CFG.gene_select.center_dim = 8
+    CFG.gene_select.expr_dim = 100
+    CFG.gene_select.work_dir = "./results/explain"
+    CFG.gene_select.gae_epochs = 2
 
     N_CELL1 = 100
     graph1 = Data(
@@ -27,19 +26,19 @@ def test_GAE_mimic():
         expr=torch.randn(N_CELL2, 100),
     )
     graph_all = Batch.from_data_list([graph1, graph2])
-    train_GAE(graph1, GENESELECT_CFG, save_dir="test_GAE")
-    train_GAE(graph_all, GENESELECT_CFG, save_dir="test_GAE_batched")
+    train_GAE(graph1, CFG.gene_select, save_dir="test_GAE")
+    train_GAE(graph_all, CFG.gene_select, save_dir="test_GAE_batched")
 
 
 def test_regress_mimic():
-    REGRESS_CFG.center_dim = 8
-    REGRESS_CFG.nbr_dim = 8
-    REGRESS_CFG.work_dir = "./results/explain"
-    REGRESS_CFG.fit.epochs = 2
+    CFG.regress.center_dim = 8
+    CFG.regress.nbr_dim = 8
+    CFG.regress.work_dir = "./results/explain"
+    CFG.regress.trainer.epochs = 2
 
     x = torch.randn(100, 8)
     y = torch.randn(100, 8)
-    train_regress(x, y, REGRESS_CFG, save_dir="test_regress")
+    train_regress(x, y, CFG.regress, save_dir="test_regress")
 
 
 if __name__ == "__main__":