MEDS-DEV 0.0.2__tar.gz → 0.0.4__tar.gz

{meds_dev-0.0.2 → meds_dev-0.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: MEDS_DEV
-Version: 0.0.2
+Version: 0.0.4
 Summary: Task configuration and helper files for the MEDS-DEV Benchmark
 Author: Nassim Oufattole
 Author-email: "Matthew B. A. McDermott" <matthew_mcdermott@hms.harvard.edu>, Ethan Steinberg <ethanid@stanford.edu>, Shalmali Joshi <shalmali.joshi@columbia.edu>, "Teya S. Bergamaschi" <teya@mit.edu>, Aleksia Kolo <aleksiak@mit.edu>, Young Sang Choi <young.sang.choi@columbia.edu>, Chao Pang <cp3016@cumc.columbia.edu>, Kamilė Stankevičiūtė <kamile.stankeviciute@cst.cam.ac.uk>, Vincent Jeanselme <vincent.jeanselme@mrc-bsu.cam.ac.uk>, Justin Xu <justin.xu@ndm.ox.ac.uk>

{meds_dev-0.0.2 → meds_dev-0.0.4}/src/MEDS_DEV/configs/_build_dataset.yaml

@@ -9,7 +9,7 @@ do_overwrite: False
 
 hydra:
   job:
-    name: "meds_dev_build_{dataset}_${now:%Y-%m-%d_%H-%M-%S}"
+    name: "meds_dev_build_dataset_${now:%Y-%m-%d_%H-%M-%S}"
   run:
     dir: "${output_dir}/.logs"
   help:

{meds_dev-0.0.2 → meds_dev-0.0.4}/src/MEDS_DEV/configs/_extract_task.yaml

@@ -3,13 +3,14 @@ defaults:
 
 dataset: ???
 dataset_dir: ???
+dataset_predicates_path: null
 task: ???
 output_dir: ???
 do_overwrite: False
 
 hydra:
   job:
-    name: "meds_dev_extract_{task}_for_{dataset}_${now:%Y-%m-%d_%H-%M-%S}"
+    name: "meds_dev_extract_task_${now:%Y-%m-%d_%H-%M-%S}"
   run:
     dir: "${output_dir}/.logs"
   help:
@@ -22,4 +23,6 @@ hydra:
 
       You can specify "dataset" to give the name of the dataset you want to build and "dataset_dir" to point
       to the extracted dataset on disk, "output_dir" to say where the final, labeled task cohort
-      should be stored on disk, and "task" to dictate which task should be extracted.
+      should be stored on disk, and "task" to dictate which task should be extracted. If you overwrite
+      "dataset_predicates_path", then it will look at that location for the predicates file, rather than in
+      the MEDS-DEV repository location. This is useful for local datasets.

{meds_dev-0.0.2 → meds_dev-0.0.4}/src/MEDS_DEV/datasets/__main__.py

@@ -1,6 +1,5 @@
 import logging
 import shutil
-import subprocess
 from pathlib import Path
 
 logger = logging.getLogger(__name__)
@@ -8,7 +7,7 @@ logger = logging.getLogger(__name__)
 import hydra
 from omegaconf import DictConfig
 
-from ..utils import temp_env
+from ..utils import run_in_env, temp_env
 from . import CFG_YAML, DATASETS
 
 
@@ -28,7 +27,7 @@ def main(cfg: DictConfig):
         output_dir.mkdir(parents=True, exist_ok=False)
 
     done_fp = output_dir / ".done"
-    if done_fp.exists():  # pragma: no cover
+    if done_fp.is_file():  # pragma: no cover
         logger.info(f"Output directory {output_dir} already exists and is marked as done.")
         return
 
@@ -37,16 +36,6 @@ def main(cfg: DictConfig):
     with temp_env(cfg, requirements) as (build_temp_dir, env):
         build_cmd = build_cmd.format(output_dir=cfg.output_dir, temp_dir=str(build_temp_dir.resolve()))
 
-        logger.info(f"Running build command: {build_cmd}")
-        build_command_out = subprocess.run(
-            build_cmd, shell=True, env=env, cwd=build_temp_dir, capture_output=True
-        )
-
-        command_errored = build_command_out.returncode != 0
-        if command_errored:
-            raise RuntimeError(
-                f"Build {cfg.dataset} command {build_cmd} failed with exit code "
-                f"{build_command_out.returncode}:\n{build_command_out.stderr.decode()}"
-            )
+        logger.info(f"Considering running build command: {build_cmd}")
+        run_in_env(build_cmd, env, cfg.output_dir, cfg.get("do_overwrite", False), cwd=build_temp_dir)
         logger.info(f"Build {cfg.dataset} command {build_cmd} completed successfully.")
-        done_fp.touch()

{meds_dev-0.0.2 → meds_dev-0.0.4}/src/MEDS_DEV/models/__main__.py

@@ -1,7 +1,6 @@
 import itertools
 import logging
 import shutil
-import subprocess
 from collections.abc import Generator
 from pathlib import Path
 
@@ -14,7 +13,7 @@ from enum import StrEnum, auto
 import hydra
 from omegaconf import DictConfig
 
-from ..utils import temp_env
+from ..utils import run_in_env, temp_env
 from . import CFG_YAML, MODELS
 
 
@@ -179,28 +178,10 @@ def main(cfg: DictConfig):
 
     with temp_env(cfg, requirements) as (temp_dir, env):
         for cmd, out_dir in model_commands(cfg, commands, model_dir):
-            done_file = out_dir / ".done"
-            if done_file.exists():
-                logger.info(f"Skipping {cmd} because {done_file} exists.")
-                continue
-
-            logger.info(f"Running model command: {cmd}")
-            command_out = subprocess.run(cmd, shell=True, env=env, cwd=temp_dir, capture_output=True)
-
-            command_errored = command_out.returncode != 0
-            if command_errored:
-                raise RuntimeError(
-                    f"{cfg.model} command {cmd} failed with exit code "
-                    f"{command_out.returncode}:\n"
-                    f"STDERR:\n{command_out.stderr.decode()}\n"
-                    f"STDOUT:\n{command_out.stdout.decode()}"
-                )
-            elif not out_dir.is_dir():
-                raise RuntimeError(
-                    f"{cfg.model} command {cmd} failed to create output directory {out_dir}.\n"
-                    f"STDERR:\n{command_out.stderr.decode()}\n"
-                    f"STDOUT:\n{command_out.stdout.decode()}"
-                )
-            else:
-                done_file.touch()
+            logger.info(f"Considering running model command: {cmd}")
+            try:
+                run_in_env(cmd, env, out_dir, cfg.get("do_overwrite", False))
+            except Exception as e:
+                raise ValueError(f"Failed to run {cfg.model} command {cmd}") from e
+
     logger.info(f"Model {cfg.model} finished successfully.")

{meds_dev-0.0.2 → meds_dev-0.0.4}/src/MEDS_DEV/tasks/__main__.py

@@ -16,11 +16,15 @@ logger = logging.getLogger(__name__)
 def main(cfg: DictConfig):
     if cfg.task not in TASKS:
         raise ValueError(f"Task {cfg.task} not currently configured")
-    if cfg.dataset not in DATASETS:
-        raise ValueError(f"Dataset {cfg.dataset} not currently configured")
 
     task_config_path = TASKS[cfg.task]["criteria_fp"]
-    dataset_predicates_path = DATASETS[cfg.dataset]["predicates"]
+    if cfg.get("dataset_predicates_path", None):
+        logger.info(f"Using provided (local) predicates path: {cfg.dataset_predicates_path}")
+        dataset_predicates_path = Path(cfg.dataset_predicates_path)
+    else:
+        if cfg.dataset not in DATASETS:
+            raise ValueError(f"Dataset {cfg.dataset} not currently configured")
+        dataset_predicates_path = DATASETS[cfg.dataset]["predicates"]
 
     if dataset_predicates_path is None:
         raise ValueError(f"Predicates not found for dataset {cfg.dataset}")

meds_dev-0.0.4/src/MEDS_DEV/utils.py

@@ -0,0 +1,131 @@
+import contextlib
+import logging
+import os
+import shutil
+import subprocess
+import sys
+import tempfile
+from pathlib import Path
+
+from omegaconf import DictConfig
+
+logger = logging.getLogger(__name__)
+
+
+def get_venv_bin_path(venv_path: str | Path) -> Path:
+    """Get the bin/Scripts directory of the virtual environment."""
+    if os.name == "nt":  # Windows
+        return Path(venv_path) / "Scripts"
+    else:  # Unix-like systems
+        return Path(venv_path) / "bin"
+
+
+@contextlib.contextmanager
+def tempdir_ctx(cfg: DictConfig) -> Path:
+    temp_dir = cfg.get("temp_dir", None)
+    if temp_dir is None:
+        with tempfile.TemporaryDirectory() as temp_dir:
+            yield Path(temp_dir)
+    else:
+        temp_dir = Path(temp_dir)
+        temp_dir.mkdir(exist_ok=True, parents=True)
+        yield temp_dir
+
+
+def install_venv(venv_path: Path, requirements: str | Path) -> Path:
+    logger.info(f"Installing requirements from {requirements} into virtual environment.")
+    subprocess.run([sys.executable, "-m", "venv", str(venv_path)], check=True)
+
+    venv_bin_path = get_venv_bin_path(venv_path)
+    venv_python = venv_bin_path / "python"
+    if not venv_python.exists():
+        raise RuntimeError(f"Virtual environment python {venv_python} does not exist!")
+
+    subprocess.run(
+        [str(venv_python), "-m", "pip", "install", "-r", str(requirements)],
+        check=True,
+    )
+
+    logger.info(f"Installed requirements from {requirements} into virtual environment.")
+    return venv_bin_path
+
+
+@contextlib.contextmanager
+def temp_env(cfg: DictConfig, requirements: str | Path | None) -> tuple[Path, dict]:
+    with tempdir_ctx(cfg) as build_temp_dir:
+        env = os.environ.copy()
+        if requirements is not None:
+            venv_path = build_temp_dir / ".venv"
+            venv_bin_path = install_venv(venv_path, requirements)
+            env["VIRTUAL_ENV"] = str(venv_path.resolve())
+            env["PATH"] = f"{str(venv_bin_path.resolve())}{os.pathsep}{env['PATH']}"
+
+        yield build_temp_dir, env
+
+
+def run_in_env(
+    cmd: str,
+    env: dict[str, str],
+    output_dir: Path | str,
+    do_overwrite: bool = False,
+    cwd: Path | str | None = None,
+) -> subprocess.CompletedProcess:
+    if type(output_dir) is str:
+        output_dir = Path(output_dir)
+
+    if do_overwrite and output_dir.exists():
+        logger.info(f"Removing existing output directory: {output_dir}")
+        shutil.rmtree(output_dir)
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    done_file = output_dir / ".done"
+    if done_file.is_file():
+        logger.info(f"Skipping {cmd} because {done_file} exists.")
+        return
+
+    script_file = output_dir / "cmd.sh"
+    script_lines = ["#!/bin/bash"]
+
+    if env.get("VIRTUAL_ENV", None) is not None:
+        script_lines.append(f"source {env['VIRTUAL_ENV']}/bin/activate")
+
+    script_lines.append(cmd)
+    script = "\n".join(script_lines)
+
+    if script_file.is_file():
+        if script_file.read_text() != script:
+            raise RuntimeError(
+                f"Script file {script_file} already exists and is different from the current script. "
+                f"Existing file:\n{script_file.read_text()}\n"
+                f"New script:\n{script}"
+                "Consider running with do_overwrite=True."
+            )
+        else:
+            logger.info(f"(Matching) script file already exists: {script_file}")
+    else:
+        script_file.write_text(script)
+
+    script_file.chmod(0o755)
+
+    logger.info(f"Running command in {script_file}:\n{script}")
+
+    runner_kwargs = {"shell": False, "env": env, "capture_output": True}
+    if cwd is not None:
+        runner_kwargs["cwd"] = cwd
+
+    command_out = subprocess.run(["bash", str(script_file.resolve())], **runner_kwargs)
+
+    command_errored = command_out.returncode != 0
+    if command_errored:
+        raise RuntimeError(
+            f"Command failed with exit code "
+            f"{command_out.returncode}:\n"
+            f"SCRIPT:\n{script}\n"
+            f"STDERR:\n{command_out.stderr.decode()}\n"
+            f"STDOUT:\n{command_out.stdout.decode()}"
+        )
+    else:
+        done_file.touch()
+
+    return command_out

{meds_dev-0.0.2 → meds_dev-0.0.4}/src/MEDS_DEV.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: MEDS_DEV
-Version: 0.0.2
+Version: 0.0.4
 Summary: Task configuration and helper files for the MEDS-DEV Benchmark
 Author: Nassim Oufattole
 Author-email: "Matthew B. A. McDermott" <matthew_mcdermott@hms.harvard.edu>, Ethan Steinberg <ethanid@stanford.edu>, Shalmali Joshi <shalmali.joshi@columbia.edu>, "Teya S. Bergamaschi" <teya@mit.edu>, Aleksia Kolo <aleksiak@mit.edu>, Young Sang Choi <young.sang.choi@columbia.edu>, Chao Pang <cp3016@cumc.columbia.edu>, Kamilė Stankevičiūtė <kamile.stankeviciute@cst.cam.ac.uk>, Vincent Jeanselme <vincent.jeanselme@mrc-bsu.cam.ac.uk>, Justin Xu <justin.xu@ndm.ox.ac.uk>

{meds_dev-0.0.2 → meds_dev-0.0.4}/tests/test_datasets.py

@@ -27,5 +27,17 @@ def test_datasets_configured(demo_dataset):
     assert demo_dataset_dir.exists(), f"Output directory not found for {dataset_name}"
     data_subdir = demo_dataset_dir / "data"
     metadata_subdir = demo_dataset_dir / "metadata"
-    assert data_subdir.exists(), f"Data directory not found for {dataset_name}"
-    assert metadata_subdir.exists(), f"Metadata directory not found for {dataset_name}"
+
+    if not data_subdir.is_dir() or not metadata_subdir.is_dir():
+        err_str = [
+            f"Either data/ or metadata/ is not a directory for {dataset_name}.",
+            f"Demo dataset directory: {demo_dataset_dir}",
+            "Directory contents:",
+        ]
+        for f in demo_dataset_dir.rglob("*"):
+            err_str.append(f"  - {f}")
+            if f.suffix == ".log":
+                err_str.append("    Log contents:")
+                err_str.append(f"    {f.read_text()}")
+
+        raise AssertionError("\n".join(err_str))

meds_dev-0.0.2/src/MEDS_DEV/utils.py

@@ -1,61 +0,0 @@
-import contextlib
-import logging
-import os
-import subprocess
-import sys
-import tempfile
-from pathlib import Path
-
-from omegaconf import DictConfig
-
-logger = logging.getLogger(__name__)
-
-
-def get_venv_bin_path(venv_path: str | Path) -> Path:
-    """Get the bin/Scripts directory of the virtual environment."""
-    if os.name == "nt":  # Windows
-        return Path(venv_path) / "Scripts"
-    else:  # Unix-like systems
-        return Path(venv_path) / "bin"
-
-
-@contextlib.contextmanager
-def tempdir_ctx(cfg: DictConfig) -> Path:
-    temp_dir = cfg.get("temp_dir", None)
-    if temp_dir is None:
-        with tempfile.TemporaryDirectory() as temp_dir:
-            yield Path(temp_dir)
-    else:
-        temp_dir = Path(temp_dir)
-        temp_dir.mkdir(exist_ok=True, parents=True)
-        yield temp_dir
-
-
-def install_venv(venv_path: Path, requirements: str | Path) -> Path:
-    logger.info(f"Installing requirements from {requirements} into virtual environment.")
-    subprocess.run([sys.executable, "-m", "venv", str(venv_path)], check=True)
-
-    venv_bin_path = get_venv_bin_path(venv_path)
-    venv_python = venv_bin_path / "python"
-    if not venv_python.exists():
-        raise RuntimeError(f"Virtual environment python {venv_python} does not exist!")
-
-    subprocess.run(
-        [str(venv_python), "-m", "pip", "install", "-r", str(requirements)],
-        check=True,
-    )
-
-    logger.info(f"Installed requirements from {requirements} into virtual environment.")
-    return venv_bin_path
-
-
-@contextlib.contextmanager
-def temp_env(cfg: DictConfig, requirements: str | Path | None) -> tuple[Path, dict]:
-    with tempdir_ctx(cfg) as build_temp_dir:
-        env = os.environ.copy()
-        if requirements is not None:
-            venv_path = build_temp_dir / ".venv"
-            venv_bin_path = install_venv(venv_path, requirements)
-            env["PATH"] = f"{str(venv_bin_path)}{os.pathsep}{env['PATH']}"
-
-        yield build_temp_dir, env