hydraflow 0.10.1__tar.gz → 0.11.0__tar.gz

{hydraflow-0.10.1 → hydraflow-0.11.0}/.gitignore

@@ -1,5 +1,5 @@
 *.db
-.coverage
+.coverage*
 .env
 .venv/
 __pycache__/

{hydraflow-0.10.1 → hydraflow-0.11.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hydraflow
-Version: 0.10.1
+Version: 0.11.0
 Summary: Hydraflow integrates Hydra and MLflow to manage and track machine learning experiments.
 Project-URL: Documentation, https://daizutabi.github.io/hydraflow/
 Project-URL: Source, https://github.com/daizutabi/hydraflow

{hydraflow-0.10.1 → hydraflow-0.11.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "hydraflow"
-version = "0.10.1"
+version = "0.11.0"
 description = "Hydraflow integrates Hydra and MLflow to manage and track machine learning experiments."
 readme = "README.md"
 license = { file = "LICENSE" }
@@ -38,6 +38,7 @@ hydraflow = "hydraflow.cli:app"
 
 [dependency-groups]
 dev = [
+  "hydra-joblib-launcher",
   "pytest-clarity",
   "pytest-cov",
   "pytest-order",

{hydraflow-0.10.1 → hydraflow-0.11.0}/src/hydraflow/cli.py

@@ -2,54 +2,53 @@
 
 from __future__ import annotations
 
-from typing import TYPE_CHECKING, Annotated
+from typing import Annotated
 
 import typer
 from rich.console import Console
 from typer import Argument, Option
 
-from hydraflow.executor.io import load_config
-
-if TYPE_CHECKING:
-    from hydraflow.executor.job import Job
+from hydraflow.executor.io import get_job, load_config
 
 app = typer.Typer(add_completion=False)
 console = Console()
 
 
-def get_job(name: str) -> Job:
-    cfg = load_config()
-    job = cfg.jobs[name]
-
-    if not job.name:
-        job.name = name
-
-    return job
-
-
 @app.command()
 def run(
     name: Annotated[str, Argument(help="Job name.", show_default=False)],
+    *,
+    dry_run: Annotated[
+        bool,
+        Option("--dry-run", help="Perform a dry run"),
+    ] = False,
 ) -> None:
     """Run a job."""
     import mlflow
 
-    from hydraflow.executor.job import multirun
+    from hydraflow.executor.job import multirun, to_text
 
     job = get_job(name)
-    mlflow.set_experiment(job.name)
-    multirun(job)
+    if dry_run:
+        typer.echo(to_text(job))
+    else:
+        mlflow.set_experiment(job.name)
+        multirun(job)
 
 
 @app.command()
 def show(
-    name: Annotated[str, Argument(help="Job name.", show_default=False)],
+    name: Annotated[str, Argument(help="Job name.", show_default=False)] = "",
 ) -> None:
-    """Show a job."""
-    from hydraflow.executor.job import show
+    """Show the hydraflow config."""
+    from omegaconf import OmegaConf
 
-    job = get_job(name)
-    show(job)
+    if name:
+        cfg = get_job(name)
+    else:
+        cfg = load_config()
+
+    typer.echo(OmegaConf.to_yaml(cfg))
 
 
 @app.callback(invoke_without_command=True)

{hydraflow-0.10.1 → hydraflow-0.11.0}/src/hydraflow/executor/conf.py

@@ -5,9 +5,9 @@ from dataclasses import dataclass, field
 
 @dataclass
 class Step:
-    args: str = ""
     batch: str = ""
-    options: str = ""
+    args: str = ""
+    configs: str = ""
 
 
 @dataclass
@@ -15,6 +15,7 @@ class Job:
     name: str = ""
     run: str = ""
     call: str = ""
+    configs: str = ""
     steps: list[Step] = field(default_factory=list)
 
 

{hydraflow-0.10.1 → hydraflow-0.11.0}/src/hydraflow/executor/io.py

@@ -3,11 +3,15 @@
 from __future__ import annotations
 
 from pathlib import Path
+from typing import TYPE_CHECKING
 
 from omegaconf import OmegaConf
 
 from .conf import HydraflowConf
 
+if TYPE_CHECKING:
+    from .job import Job
+
 
 def find_config_file() -> Path | None:
     """Find the hydraflow config file."""
@@ -32,3 +36,14 @@ def load_config() -> HydraflowConf:
     cfg = OmegaConf.load(path)
 
     return OmegaConf.merge(schema, cfg)  # type: ignore
+
+
+def get_job(name: str) -> Job:
+    """Get a job from the config."""
+    cfg = load_config()
+    job = cfg.jobs[name]
+
+    if not job.name:
+        job.name = name
+
+    return job

{hydraflow-0.10.1 → hydraflow-0.11.0}/src/hydraflow/executor/job.py

@@ -12,7 +12,7 @@ The module supports two execution modes:
 2. Python function calls
 
 Each job can consist of multiple steps, and each step can have its own
-arguments and options that will be expanded into multiple runs.
+arguments and configurations that will be expanded into multiple runs.
 """
 
 from __future__ import annotations
@@ -20,6 +20,7 @@ from __future__ import annotations
 import importlib
 import shlex
 import subprocess
+import sys
 from subprocess import CalledProcessError
 from typing import TYPE_CHECKING
 
@@ -30,27 +31,27 @@ from .parser import collect, expand
 if TYPE_CHECKING:
     from collections.abc import Iterator
 
-    from .conf import Job, Step
+    from .conf import Job
 
 
-def iter_args(step: Step) -> Iterator[list[str]]:
+def iter_args(batch: str, args: str) -> Iterator[list[str]]:
     """Iterate over combinations generated from parsed arguments.
 
     Generate all possible combinations of arguments by parsing and
     expanding each one, yielding them as an iterator.
 
     Args:
-        step (Step): The step to parse.
+        batch (str): The batch to parse.
+        args (str): The arguments to parse.
 
     Yields:
         list[str]: a list of the parsed argument combinations.
 
     """
-    args = collect(step.args)
-    options = [o for o in step.options.split(" ") if o]
+    args_ = collect(args)
 
-    for batch in expand(step.batch):
-        yield [*options, *sorted([*batch, *args])]
+    for batch_ in expand(batch):
+        yield [*batch_, *args_]
 
 
 def iter_batches(job: Job) -> Iterator[list[str]]:
@@ -68,11 +69,14 @@ def iter_batches(job: Job) -> Iterator[list[str]]:
 
     """
     job_name = f"hydra.job.name={job.name}"
+    job_configs = shlex.split(job.configs)
 
     for step in job.steps:
-        for args in iter_args(step):
+        configs = shlex.split(step.configs) or job_configs
+
+        for args in iter_args(step.batch, step.args):
             sweep_dir = f"hydra.sweep.dir=multirun/{ulid.ULID()}"
-            yield ["--multirun", sweep_dir, job_name, *args]
+            yield ["--multirun", *args, job_name, sweep_dir, *configs]
 
 
 def multirun(job: Job) -> None:
@@ -99,6 +103,8 @@ def multirun(job: Job) -> None:
 
     if job.run:
         base_cmds = shlex.split(job.run)
+        if base_cmds[0] == "python" and sys.platform == "win32":
+            base_cmds[0] = sys.executable
 
         for args in it:
             cmds = [*base_cmds, *args]
@@ -132,24 +138,31 @@ def multirun(job: Job) -> None:
                 raise RuntimeError(msg) from e
 
 
-def show(job: Job) -> None:
-    """Show the job configuration.
+def to_text(job: Job) -> str:
+    """Convert the job configuration to a string.
 
-    This function shows the job configuration for a given job.
+    This function returns the job configuration for a given job.
 
     Args:
         job (Job): The job configuration to show.
 
+    Returns:
+        str: The job configuration.
+
     """
+    text = ""
+
     it = iter_batches(job)
 
     if job.run:
         base_cmds = shlex.split(job.run)
         for args in it:
             cmds = " ".join([*base_cmds, *args])
-            print(cmds)  # noqa: T201
+            text += f"{cmds}\n"
 
     elif job.call:
-        print(f"call: {job.call}")  # noqa: T201
+        text = f"call: {job.call}\n"
         for args in it:
-            print(f"args: {args}")  # noqa: T201
+            text += f"args: {args}\n"
+
+    return text.rstrip()

{hydraflow-0.10.1 → hydraflow-0.11.0}/tests/cli/app.py

@@ -1,5 +1,7 @@
 from __future__ import annotations
 
+import logging
+import time
 from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
@@ -8,6 +10,8 @@ import hydraflow
 if TYPE_CHECKING:
     from mlflow.entities import Run
 
+log = logging.getLogger(__name__)
+
 
 @dataclass
 class Config:
@@ -17,7 +21,9 @@ class Config:
 
 @hydraflow.main(Config)
 def app(run: Run, cfg: Config):
-    pass
+    log.info("start")
+    time.sleep(0.2)
+    log.info("end")
 
 
 if __name__ == "__main__":

hydraflow-0.11.0/tests/cli/hydraflow.yaml

@@ -0,0 +1,22 @@
+jobs:
+  args:
+    run: python app.py
+    steps:
+      - args: count=1:3 name=a,b
+      - args: count=4:6 name=c,d
+  batch:
+    run: python app.py
+    steps:
+      - batch: name=a,b
+        args: count=1,2
+      - batch: name=c,d|e,f
+        args: count=100
+  parallel:
+    run: python app.py
+    configs: hydra/launcher=joblib hydra.launcher.n_jobs=2
+    steps:
+      - batch: name=a
+        args: count=1:4
+      - batch: name=b
+        args: count=11:14
+        configs: hydra/launcher=joblib hydra.launcher.n_jobs=4

hydraflow-0.11.0/tests/cli/test_run.py

@@ -0,0 +1,63 @@
+import pytest
+from typer.testing import CliRunner
+
+import hydraflow
+from hydraflow.cli import app
+
+runner = CliRunner()
+
+
+def test_run_args_dry_run():
+    result = runner.invoke(app, ["run", "args", "--dry-run"])
+    assert result.exit_code == 0
+    out = result.stdout
+    assert "hydra.job.name=args" in out
+    assert "count=1,2,3 name=a,b" in out
+    assert "count=4,5,6 name=c,d" in out
+
+
+def test_run_batch_dry_run():
+    result = runner.invoke(app, ["run", "batch", "--dry-run"])
+    assert result.exit_code == 0
+    out = result.stdout
+    assert "name=a count=1,2" in out
+    assert "name=b count=1,2" in out
+    assert "name=c,d count=100" in out
+    assert "name=e,f count=100" in out
+    assert "hydra.job.name=batch" in out
+
+
+def test_run_parallel_dry_run():
+    result = runner.invoke(app, ["run", "parallel", "--dry-run"])
+    assert result.exit_code == 0
+    out = result.stdout
+    lines = out.splitlines()
+    assert len(lines) == 2
+    assert "count=1,2,3,4" in lines[0]
+    assert "hydra.launcher.n_jobs=2" in lines[0]
+    assert "count=11,12,13,14" in lines[1]
+    assert "hydra.launcher.n_jobs=4" in lines[1]
+
+
+@pytest.mark.xdist_group(name="group1")
+def test_run_args():
+    result = runner.invoke(app, ["run", "args"])
+    assert result.exit_code == 0
+    run_ids = hydraflow.list_run_ids("args")
+    assert len(run_ids) == 12
+
+
+@pytest.mark.xdist_group(name="group2")
+def test_run_batch():
+    result = runner.invoke(app, ["run", "batch"])
+    assert result.exit_code == 0
+    run_ids = hydraflow.list_run_ids("batch")
+    assert len(run_ids) == 8
+
+
+@pytest.mark.xdist_group(name="group3")
+def test_run_parallel():
+    result = runner.invoke(app, ["run", "parallel"])
+    assert result.exit_code == 0
+    run_ids = hydraflow.list_run_ids("parallel")
+    assert len(run_ids) == 8

hydraflow-0.11.0/tests/cli/test_show.py

@@ -0,0 +1,18 @@
+from typer.testing import CliRunner
+
+from hydraflow.cli import app
+
+runner = CliRunner()
+
+
+def test_show():
+    result = runner.invoke(app, ["show"])
+    assert result.exit_code == 0
+    assert "jobs:\n" in result.stdout
+    assert "  args:\n" in result.stdout
+
+
+def test_show_job():
+    result = runner.invoke(app, ["show", "args"])
+    assert result.exit_code == 0
+    assert "name: args\n" in result.stdout

{hydraflow-0.10.1 → hydraflow-0.11.0}/tests/executor/echo.py

@@ -4,7 +4,7 @@ from pathlib import Path
 
 def main():
     path = Path(sys.argv[1])
-    arg = " ".join(sys.argv[-2:])
+    arg = " ".join(sys.argv[3:5])
 
     if not path.exists():
         path.write_text(arg)

{hydraflow-0.10.1 → hydraflow-0.11.0}/tests/executor/test_conf.py

@@ -30,5 +30,5 @@ def test_job(config):
 
 
 def test_step(config):
-    cfg = config("jobs:\n  a:\n    steps:\n      - options: --opt1 --opt2\n")
-    assert cfg.jobs["a"].steps[0].options == "--opt1 --opt2"
+    cfg = config("jobs:\n  a:\n    steps:\n      - configs: --opt1 --opt2\n")
+    assert cfg.jobs["a"].steps[0].configs == "--opt1 --opt2"

{hydraflow-0.10.1 → hydraflow-0.11.0}/tests/executor/test_job.py

@@ -9,36 +9,25 @@ from hydraflow.executor.conf import Job, Step
 def test_iter_args():
     from hydraflow.executor.job import iter_args
 
-    step = Step(args="a=1:3", batch="b=3,4 c=5,6")
-    it = iter_args(step)
-    assert next(it) == ["a=1,2,3", "b=3", "c=5"]
-    assert next(it) == ["a=1,2,3", "b=3", "c=6"]
-    assert next(it) == ["a=1,2,3", "b=4", "c=5"]
-    assert next(it) == ["a=1,2,3", "b=4", "c=6"]
+    it = iter_args("b=3,4 c=5,6", "a=1:3")
+    assert next(it) == ["b=3", "c=5", "a=1,2,3"]
+    assert next(it) == ["b=3", "c=6", "a=1,2,3"]
+    assert next(it) == ["b=4", "c=5", "a=1,2,3"]
+    assert next(it) == ["b=4", "c=6", "a=1,2,3"]
 
 
 def test_iter_args_pipe():
     from hydraflow.executor.job import iter_args
 
-    step = Step(args="a=1:3", batch="b=3,4|c=5:7")
-    it = iter_args(step)
-    assert next(it) == ["a=1,2,3", "b=3,4"]
-    assert next(it) == ["a=1,2,3", "c=5,6,7"]
-
-
-def test_iter_args_with_options():
-    from hydraflow.executor.job import iter_args
-
-    step = Step(args="a=1:3", batch="b=3,4", options="--opt1 --opt2")
-    it = iter_args(step)
-    assert next(it) == ["--opt1", "--opt2", "a=1,2,3", "b=3"]
-    assert next(it) == ["--opt1", "--opt2", "a=1,2,3", "b=4"]
+    it = iter_args("b=3,4|c=5:7", "a=1:3")
+    assert next(it) == ["b=3,4", "a=1,2,3"]
+    assert next(it) == ["c=5,6,7", "a=1,2,3"]
 
 
 @pytest.fixture
 def job():
-    s1 = Step(args="a=1:2", batch="b=5,6")
-    s2 = Step(args="a=3:4", batch="c=7,8")
+    s1 = Step(batch="b=5,6", args="a=1:2")
+    s2 = Step(batch="c=7,8", args="a=3:4")
     return Job(name="test", steps=[s1, s2])
 
 
@@ -50,17 +39,17 @@ def batches(job: Job):
 
 
 def test_sweep_dir(batches):
-    assert all(x[1].startswith("hydra.sweep.dir=multirun/") for x in batches)
-    assert all(len(x[1].split("/")[-1]) == 26 for x in batches)
+    assert all(x[-1].startswith("hydra.sweep.dir=multirun/") for x in batches)
+    assert all(len(x[-1].split("/")[-1]) == 26 for x in batches)
 
 
 def test_job_name(batches):
-    assert all(x[2].startswith("hydra.job.name=test") for x in batches)
+    assert all(x[-2].startswith("hydra.job.name=test") for x in batches)
 
 
 @pytest.mark.parametrize(("i", "x"), [(0, "b=5"), (1, "b=6"), (2, "c=7"), (3, "c=8")])
 def test_batch_args(batches, i, x):
-    assert batches[i][-1] == x
+    assert batches[i][1] == x
 
 
 @pytest.mark.parametrize(
@@ -68,28 +57,24 @@ def test_batch_args(batches, i, x):
     [(0, "a=1,2"), (1, "a=1,2"), (2, "a=3,4"), (3, "a=3,4")],
 )
 def test_sweep_args(batches, i, x):
-    assert batches[i][-2] == x
+    assert batches[i][-3] == x
 
 
-@pytest.mark.skipif(
-    sys.platform == "win32",
-    reason="Windows does not support this test",
-)
 def test_multirun_run(job: Job, tmp_path: Path):
     from hydraflow.executor.job import multirun
 
     path = tmp_path / "output.txt"
     file = Path(__file__).parent / "echo.py"
 
-    job.run = f"{sys.executable} {file.as_posix()} {path.as_posix()}"
+    job.run = f"python {file.as_posix()} {path.as_posix()}"
     multirun(job)
-    assert path.read_text() == "a=1,2 b=5 a=1,2 b=6 a=3,4 c=7 a=3,4 c=8"
+    assert path.read_text() == "b=5 a=1,2 b=6 a=1,2 c=7 a=3,4 c=8 a=3,4"
 
 
 def test_multirun_run_error(job: Job):
     from hydraflow.executor.job import multirun
 
-    job.run = "false"
+    job.run = "cmd /c exit 1" if sys.platform == "win32" else "false"
     with pytest.raises(RuntimeError):
         multirun(job)
 
@@ -100,8 +85,8 @@ def test_multirun_call(job: Job, capsys: pytest.CaptureFixture):
     job.call = "typer.echo"
     multirun(job)
     out, _ = capsys.readouterr()
-    assert "'a=1,2', 'b=5'" in out
-    assert "'a=3,4', 'c=8'" in out
+    assert "'b=5', 'a=1,2'" in out
+    assert "'c=8', 'a=3,4'" in out
 
 
 def test_multirun_call_args(job: Job, capsys: pytest.CaptureFixture):
@@ -137,11 +122,11 @@ def test_multirun_call_not_found(job: Job):
         multirun(job)
 
 
-def test_show(job: Job, capsys):
-    from hydraflow.executor.job import show
+def test_to_text(job: Job):
+    from hydraflow.executor.job import to_text
 
     job.call = "typer.echo"
-    show(job)
-    out, _ = capsys.readouterr()
-    assert "call: typer.echo\n" in out
-    assert "'hydra.job.name=test', 'a=3,4', 'c=8']" in out
+    text = to_text(job)
+    assert "call: typer.echo\n" in text
+    assert "'b=5', 'a=1,2', 'hydra.job.name=test'" in text
+    assert "'c=8', 'a=3,4', 'hydra.job.name=test'" in text

hydraflow-0.10.1/tests/cli/hydraflow.yaml

@@ -1,13 +0,0 @@
-jobs:
-  args:
-    run: python app.py
-    steps:
-      - args: count=1:3 name=a,b
-      - args: count=4:6 name=c,d
-  batch:
-    run: python app.py
-    steps:
-      - args: count=1,2
-        batch: name=a,b
-      - args: count=100
-        batch: name=c,d|e,f

hydraflow-0.10.1/tests/cli/test_run.py

@@ -1,33 +0,0 @@
-import sys
-
-import pytest
-from typer.testing import CliRunner
-
-import hydraflow
-from hydraflow.cli import app
-
-pytestmark = pytest.mark.xdist_group(name="group1")
-
-runner = CliRunner()
-
-
-@pytest.mark.skipif(
-    sys.platform == "win32",
-    reason="Windows does not support this test",
-)
-def test_run_args():
-    result = runner.invoke(app, ["run", "args"])
-    assert result.exit_code == 0
-    run_ids = hydraflow.list_run_ids("args")
-    assert len(run_ids) == 12
-
-
-@pytest.mark.skipif(
-    sys.platform == "win32",
-    reason="Windows does not support this test",
-)
-def test_run_batch():
-    result = runner.invoke(app, ["run", "batch"])
-    assert result.exit_code == 0
-    run_ids = hydraflow.list_run_ids("batch")
-    assert len(run_ids) == 8

hydraflow-0.10.1/tests/cli/test_show.py

@@ -1,23 +0,0 @@
-from typer.testing import CliRunner
-
-from hydraflow.cli import app
-
-runner = CliRunner()
-
-
-def test_show_args():
-    result = runner.invoke(app, ["show", "args"])
-    assert result.exit_code == 0
-    assert "hydra.job.name=args" in result.stdout
-    assert "count=1,2,3 name=a,b" in result.stdout
-    assert "count=4,5,6 name=c,d" in result.stdout
-
-
-def test_show_batch():
-    result = runner.invoke(app, ["show", "batch"])
-    assert result.exit_code == 0
-    assert "hydra.job.name=batch" in result.stdout
-    assert "count=1,2 name=a" in result.stdout
-    assert "count=1,2 name=b" in result.stdout
-    assert "count=100 name=c,d" in result.stdout
-    assert "count=100 name=e,f" in result.stdout