expops 0.1.3__py3-none-any.whl

mlops/templates/sklearn-basic/charts/plot_metrics.py

@@ -0,0 +1,85 @@
+from __future__ import annotations
+
+import json
+from pathlib import Path
+from typing import Any, Dict, Tuple
+
+def _series(metric: Any) -> Tuple[list[int], list[float]]:
+    """Convert {"1": 0.1, "2": 0.2, ...} into ([1,2], [0.1,0.2])."""
+    if not isinstance(metric, dict):
+        return ([], [])
+    pts: list[tuple[int, float]] = []
+    for k, v in metric.items():
+        try:
+            step = int(k)
+            val = float(v)
+        except Exception:
+            continue
+        pts.append((step, val))
+    pts.sort(key=lambda x: x[0])
+    return ([p[0] for p in pts], [p[1] for p in pts])
+
+
+def _last(metric: Any) -> float | None:
+    xs, ys = _series(metric)
+    return ys[-1] if ys else None
+
+
+def plot_metrics(metrics: Dict[str, Any], ctx: Any) -> None:
+    """Chart entrypoint invoked by ExpOps chart runner.
+
+    - When a KV backend is configured (Redis/Firestore), `metrics` contains probe metrics.
+    - When no backend is configured, this template falls back to a local JSON written by `models/model.py`.
+    """
+    import matplotlib.pyplot as plt  # type: ignore
+
+    train_block = metrics.get("train") if isinstance(metrics, dict) else None
+    eval_block = metrics.get("eval") if isinstance(metrics, dict) else None
+
+    train_block = train_block if isinstance(train_block, dict) else {}
+    eval_block = eval_block if isinstance(eval_block, dict) else {}
+
+    # 1) Training loss curve
+    xs, ys = _series(train_block.get("loss"))
+    fig, ax = plt.subplots(figsize=(6, 3))
+    if ys:
+        ax.plot(xs, ys, marker="o", linewidth=1.5)
+        ax.set_xlabel("epoch")
+        ax.set_ylabel("loss")
+    else:
+        ax.text(0.5, 0.5, "No loss metrics found", ha="center", va="center", transform=ax.transAxes)
+        ax.set_xticks([])
+        ax.set_yticks([])
+    ax.set_title("Training loss")
+    ax.grid(True, alpha=0.25)
+    fig.tight_layout()
+    ctx.savefig("training_loss.png", fig=fig)
+    plt.close(fig)
+
+    # 2) Accuracy bars (final train vs eval)
+    train_acc = _last(train_block.get("accuracy"))
+    eval_acc = _last(eval_block.get("accuracy"))
+    labels: list[str] = []
+    values: list[float] = []
+    if train_acc is not None:
+        labels.append("train")
+        values.append(float(train_acc))
+    if eval_acc is not None:
+        labels.append("eval")
+        values.append(float(eval_acc))
+
+    fig, ax = plt.subplots(figsize=(4, 3))
+    if values:
+        ax.bar(labels, values)
+        ax.set_ylim(0.0, 1.0)
+        ax.set_ylabel("accuracy")
+    else:
+        ax.text(0.5, 0.5, "No accuracy metrics found", ha="center", va="center", transform=ax.transAxes)
+        ax.set_xticks([])
+        ax.set_yticks([])
+    ax.set_title("Accuracy")
+    ax.grid(True, axis="y", alpha=0.25)
+    fig.tight_layout()
+    ctx.savefig("accuracy.png", fig=fig)
+    plt.close(fig)
+

mlops/templates/sklearn-basic/charts/requirements.txt

@@ -0,0 +1,3 @@
+# Chart/reporting dependencies for this template.
+matplotlib
+

mlops/templates/sklearn-basic/configs/project_config.yaml

@@ -0,0 +1,64 @@
+metadata:
+  name: "sklearn-basic"
+  description: "Template project: custom pipeline (sklearn inside) + simple reporting chart"
+  version: "1.0.0"
+
+environment:
+  venv:
+    name: "sklearn-basic-env"
+    requirements_file: "projects/sklearn-basic/requirements.txt"
+    reporting:
+      name: "sklearn-basic-env-reporting"
+      requirements_file: "projects/sklearn-basic/charts/requirements.txt"
+
+reproducibility:
+  random_seed: 42
+
+model:
+  framework: "custom"
+  language: "python"
+  name: "sklearn_basic_custom"
+  version: "1.0.0"
+  parameters:
+    custom_script_path: "projects/sklearn-basic/models/model.py"
+    cache:
+      backend:
+        # Local-first default
+        type: memory
+        # --- GCP example (Firestore + Pub/Sub) ---
+        # type: gcp
+        # gcp_project: your-gcp-project-id
+        # credentials_json: keys/firestore.json
+        # emulator_host: 127.0.0.1:8080
+      # Optional object store for cache artifacts.
+      # object_store:
+      #   type: gcs
+      #   bucket: your-gcs-bucket-name
+      #   prefix: projects/{{PROJECT_ID}}/cache/steps
+    executor:
+      n_workers: 2
+    pipeline:
+      process_adjlist: |
+        train_model evaluate_model
+        evaluate_model plot_metrics
+      processes:
+        - name: "train_model"
+          description: "Train a tiny classifier and log loss/accuracy"
+          code_function: "train_model"
+
+        - name: "evaluate_model"
+          description: "Evaluate the trained model and log accuracy"
+          code_function: "evaluate_model"
+
+        - name: "plot_metrics"
+          type: chart
+          description: "Render a basic PNG report from recorded metrics"
+
+reporting:
+  static_entrypoint: "projects/sklearn-basic/charts/plot_metrics.py"
+  charts:
+    - name: "plot_metrics"
+      probe_paths:
+        train: "train_model"
+        eval: "evaluate_model"
+

mlops/templates/sklearn-basic/data/train.csv

@@ -0,0 +1,14 @@
+feature_1,feature_2,label
+0.10,0.20,0
+0.20,0.10,0
+0.15,0.25,0
+0.30,0.20,0
+0.90,1.10,1
+1.10,0.90,1
+1.00,1.00,1
+0.85,1.20,1
+0.40,0.35,0
+0.55,0.45,0
+0.70,0.80,1
+0.75,0.65,1
+

mlops/templates/sklearn-basic/models/model.py

@@ -0,0 +1,62 @@
+from __future__ import annotations
+
+from pathlib import Path
+
+import pandas as pd
+from sklearn.linear_model import LogisticRegression
+from sklearn.metrics import accuracy_score
+
+from mlops.core import process, log_metric
+
+
+def _load_xy(csv_path: str | Path):
+    df = pd.read_csv(Path(csv_path))
+    y = df.pop("label").values
+    x = df.values
+    return x, y
+
+
+def _training_csv_path() -> Path:
+    return Path(__file__).resolve().parents[1] / "data" / "train.csv"
+
+
+def _validation_csv_path() -> Path:
+    return _training_csv_path()
+
+
+@process(description="Train a tiny classifier")
+def train_model(data):
+    train_path = _training_csv_path()
+    if not train_path.exists():
+        raise FileNotFoundError(
+            f"Training CSV not found at {train_path}. "
+            "Expected the template dataset at projects/<project_id>/data/train.csv."
+        )
+    x, y = _load_xy(train_path)
+    model = LogisticRegression(max_iter=200)
+    model.fit(x, y)
+
+    train_acc = float(accuracy_score(y, model.predict(x)))
+    log_metric("accuracy", train_acc, step=1)
+
+    return {"model": model, "train_accuracy": train_acc}
+
+
+@process(description="Evaluate the model")
+def evaluate_model(data):
+    model = (data or {}).get("train_model", {}).get("model")
+    if model is None:
+        raise ValueError("Missing upstream model. Expected data['train_model']['model'].")
+
+    val_path = _validation_csv_path()
+    if not val_path.exists():
+        raise FileNotFoundError(
+            f"Validation CSV not found at {val_path}. "
+            "Expected the template dataset at projects/<project_id>/data/train.csv."
+        )
+    x, y = _load_xy(val_path)
+    eval_acc = float(accuracy_score(y, model.predict(x)))
+    log_metric("accuracy", eval_acc, step=1)
+
+    return {"evaluation_accuracy": eval_acc}
+

mlops/templates/sklearn-basic/requirements.txt

@@ -0,0 +1,10 @@
+# Project runtime dependencies for this template.
+#
+# Note: ExpOps (`expops`) itself is installed into the project venv automatically
+# when you run `mlops run <project-id>`.
+
+pandas
+numpy
+scikit-learn
+joblib
+

mlops/web/__init__.py

@@ -0,0 +1,3 @@
+"""Web app package for simple local UI (API + static frontend)."""
+
+