필사 모드: Kubeflow Pipelines v2 실전 가이드 — KFP SDK로 ML 파이프라인 구축하기

pip install kfp==2.7.0

# Kubeflow Pipelines 백엔드 설치 (Kubernetes)
kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/env/platform-agnostic?ref=2.2.0"

# 포트포워딩
kubectl port-forward svc/ml-pipeline-ui -n kubeflow 8080:80

# 1. Component: 파이프라인의 단위 작업 (Python 함수)
# 2. Pipeline: Component들의 DAG(방향 비순환 그래프)
# 3. Artifact: 입출력 데이터 (Dataset, Model, Metrics 등)
# 4. Run: 파이프라인의 한 번 실행
# 5. Experiment: Run들의 논리적 그룹

from kfp import dsl
from kfp.dsl import (
    Dataset, Input, Output, Model, Metrics,
    ClassificationMetrics, component
)


@dsl.component(
    base_image="python:3.11-slim",
    packages_to_install=["pandas==2.1.4", "scikit-learn==1.4.0"]
)
def load_data(
    dataset_url: str,
    output_dataset: Output[Dataset]
):
    """데이터 로드 컴포넌트"""
    import pandas as pd

    df = pd.read_csv(dataset_url)
    print(f"Loaded {len(df)} rows")

    # Output artifact에 저장
    df.to_csv(output_dataset.path, index=False)
    output_dataset.metadata["num_rows"] = len(df)
    output_dataset.metadata["num_columns"] = len(df.columns)


@dsl.component(
    base_image="python:3.11-slim",
    packages_to_install=["pandas==2.1.4", "scikit-learn==1.4.0"]
)
def preprocess_data(
    input_dataset: Input[Dataset],
    train_dataset: Output[Dataset],
    test_dataset: Output[Dataset],
    test_size: float = 0.2
):
    """데이터 전처리 및 분할"""
    import pandas as pd
    from sklearn.model_selection import train_test_split

    df = pd.read_csv(input_dataset.path)

    # 전처리
    df = df.dropna()
    df = df.drop_duplicates()

    # 분할
    train_df, test_df = train_test_split(df, test_size=test_size, random_state=42)

    train_df.to_csv(train_dataset.path, index=False)
    test_df.to_csv(test_dataset.path, index=False)

    train_dataset.metadata["num_rows"] = len(train_df)
    test_dataset.metadata["num_rows"] = len(test_df)


@dsl.component(
    base_image="python:3.11-slim",
    packages_to_install=[
        "pandas==2.1.4", "scikit-learn==1.4.0",
        "joblib==1.3.2", "xgboost==2.0.3"
    ]
)
def train_model(
    train_dataset: Input[Dataset],
    model_output: Output[Model],
    metrics_output: Output[Metrics],
    n_estimators: int = 100,
    max_depth: int = 6,
    learning_rate: float = 0.1
):
    """모델 학습"""
    import pandas as pd
    import joblib
    from xgboost import XGBClassifier
    from sklearn.model_selection import cross_val_score

    df = pd.read_csv(train_dataset.path)
    X = df.drop("target", axis=1)
    y = df["target"]

    # 학습
    model = XGBClassifier(
        n_estimators=n_estimators,
        max_depth=max_depth,
        learning_rate=learning_rate,
        random_state=42
    )
    model.fit(X, y)

    # 교차 검증
    cv_scores = cross_val_score(model, X, y, cv=5, scoring="accuracy")

    # 모델 저장
    joblib.dump(model, model_output.path)
    model_output.metadata["framework"] = "xgboost"
    model_output.metadata["n_estimators"] = n_estimators

    # 메트릭 기록
    metrics_output.log_metric("cv_accuracy_mean", float(cv_scores.mean()))
    metrics_output.log_metric("cv_accuracy_std", float(cv_scores.std()))
    metrics_output.log_metric("n_estimators", n_estimators)


@dsl.component(
    base_image="python:3.11-slim",
    packages_to_install=[
        "pandas==2.1.4", "scikit-learn==1.4.0",
        "joblib==1.3.2", "xgboost==2.0.3"
    ]
)
def evaluate_model(
    test_dataset: Input[Dataset],
    model_input: Input[Model],
    metrics_output: Output[ClassificationMetrics],
    eval_metrics: Output[Metrics]
) -> float:
    """모델 평가"""
    import pandas as pd
    import joblib
    from sklearn.metrics import accuracy_score, classification_report

    df = pd.read_csv(test_dataset.path)
    X = df.drop("target", axis=1)
    y = df["target"]

    model = joblib.load(model_input.path)
    y_pred = model.predict(X)
    y_prob = model.predict_proba(X)

    accuracy = accuracy_score(y, y_pred)

    # Classification metrics (Confusion Matrix 시각화)
    metrics_output.log_confusion_matrix(
        categories=["Class 0", "Class 1"],
        matrix=[[int(sum((y == 0) & (y_pred == 0))), int(sum((y == 0) & (y_pred == 1)))],
                [int(sum((y == 1) & (y_pred == 0))), int(sum((y == 1) & (y_pred == 1)))]]
    )

    eval_metrics.log_metric("test_accuracy", accuracy)

    return accuracy

@dsl.component(
    base_image="pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime",
    packages_to_install=["transformers==4.37.0", "datasets==2.16.0"]
)
def finetune_llm(
    model_name: str,
    train_dataset: Input[Dataset],
    output_model: Output[Model],
    epochs: int = 3,
    batch_size: int = 8
):
    """LLM 파인튜닝 (GPU 사용)"""
    from transformers import AutoModelForSequenceClassification, Trainer
    # ... 학습 코드
    pass

@dsl.pipeline(
    name="ML Training Pipeline",
    description="데이터 로드 → 전처리 → 학습 → 평가 파이프라인"
)
def ml_training_pipeline(
    dataset_url: str = "https://example.com/data.csv",
    test_size: float = 0.2,
    n_estimators: int = 100,
    max_depth: int = 6,
    learning_rate: float = 0.1,
    accuracy_threshold: float = 0.85
):
    # Step 1: 데이터 로드
    load_task = load_data(dataset_url=dataset_url)

    # Step 2: 전처리 (load_task 완료 후 실행)
    preprocess_task = preprocess_data(
        input_dataset=load_task.outputs["output_dataset"],
        test_size=test_size
    )

    # Step 3: 모델 학습
    train_task = train_model(
        train_dataset=preprocess_task.outputs["train_dataset"],
        n_estimators=n_estimators,
        max_depth=max_depth,
        learning_rate=learning_rate
    )
    # GPU 리소스 설정
    train_task.set_cpu_limit("4")
    train_task.set_memory_limit("8Gi")

    # Step 4: 평가
    eval_task = evaluate_model(
        test_dataset=preprocess_task.outputs["test_dataset"],
        model_input=train_task.outputs["model_output"]
    )

    # Step 5: 조건부 배포
    with dsl.If(eval_task.output >= accuracy_threshold):
        deploy_task = deploy_model(
            model_input=train_task.outputs["model_output"],
            accuracy=eval_task.output
        )


@dsl.component(base_image="python:3.11-slim")
def deploy_model(
    model_input: Input[Model],
    accuracy: float
):
    """모델 배포 (조건 충족 시)"""
    print(f"Deploying model with accuracy: {accuracy:.4f}")
    print(f"Model path: {model_input.path}")
    # 실제 배포 로직 (K8s Serving, BentoML 등)

from kfp import compiler
from kfp.client import Client

# 1. YAML로 컴파일
compiler.Compiler().compile(
    pipeline_func=ml_training_pipeline,
    package_path="ml_pipeline.yaml"
)

# 2. KFP 서버에 제출
client = Client(host="http://localhost:8080")

# Experiment 생성
experiment = client.create_experiment(name="ml-experiments")

# Run 실행
run = client.create_run_from_pipeline_func(
    ml_training_pipeline,
    experiment_name="ml-experiments",
    run_name="training-run-001",
    arguments={
        "dataset_url": "gs://my-bucket/data.csv",
        "n_estimators": 200,
        "max_depth": 8,
        "accuracy_threshold": 0.90
    }
)

print(f"Run ID: {run.run_id}")
print(f"Run URL: http://localhost:8080/#/runs/details/{run.run_id}")

# 매일 새벽 2시에 실행
client.create_recurring_run(
    experiment_id=experiment.experiment_id,
    job_name="daily-retraining",
    pipeline_func=ml_training_pipeline,
    cron_expression="0 2 * * *",
    max_concurrency=1,
    arguments={
        "dataset_url": "gs://my-bucket/latest-data.csv",
        "accuracy_threshold": 0.85
    }
)

@dsl.pipeline(name="Hyperparameter Search")
def hp_search_pipeline():
    # 하이퍼파라미터 조합 정의
    hp_configs = [
        {"n_estimators": 100, "max_depth": 4, "lr": 0.1},
        {"n_estimators": 200, "max_depth": 6, "lr": 0.05},
        {"n_estimators": 300, "max_depth": 8, "lr": 0.01},
    ]

    # 병렬 학습
    with dsl.ParallelFor(hp_configs) as config:
        train_task = train_model(
            train_dataset=load_task.outputs["output_dataset"],
            n_estimators=config.n_estimators,
            max_depth=config.max_depth,
            learning_rate=config.lr
        )

# 컴포넌트 레벨에서 캐싱 비활성화
load_task = load_data(dataset_url=dataset_url)
load_task.set_caching_options(False)  # 항상 새로 실행

# 파이프라인 레벨에서 캐싱 설정
run = client.create_run_from_pipeline_func(
    ml_training_pipeline,
    enable_caching=True  # 동일 입력이면 캐시 사용
)

@dsl.component(base_image="python:3.11-slim")
def process_large_data(output_data: Output[Dataset]):
    """대용량 데이터 처리"""
    pass

# PVC 마운트
process_task = process_large_data()
process_task.add_pvolumes({
    "/mnt/data": dsl.PipelineVolume(pvc="data-pvc")
})

# .github/workflows/ml-pipeline.yml
name: ML Pipeline CI/CD

on:
  push:
    branches: [main]
    paths:
      - 'pipelines/**'
      - 'components/**'

jobs:
  deploy-pipeline:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.11'

      - name: Install dependencies
        run: pip install kfp==2.7.0

      - name: Compile pipeline
        run: python pipelines/compile.py

      - name: Upload and run pipeline
        env:
          KFP_HOST: ${{ secrets.KFP_HOST }}
        run: |
          python -c "
          from kfp.client import Client
          client = Client(host='$KFP_HOST')
          client.upload_pipeline(
            pipeline_package_path='ml_pipeline.yaml',
            pipeline_name='ml-training-v2',
            description='Automated ML training pipeline'
          )
          "

from kfp import compiler

compiler.Compiler().compile(
    pipeline_func=ml_training_pipeline,
    package_path="ml_pipeline.yaml",  # 문서 표현 그대로 "출력 YAML 파일 경로"
    pipeline_name="ml-training",
    pipeline_display_name="ML Training Pipeline",
    pipeline_parameters={"n_estimators": 200},
    type_check=True,
)
# 나머지 인자: kubernetes_manifest_options, kubernetes_manifest_format
# 주의: kfp_package_path 는 compile() 이 아니라 @dsl.component 의 인자다

kfp dsl compile --py my_pipeline.py --output my_pipeline.yaml

from kfp.client import Client

client = Client(host="http://localhost:8080", namespace="kubeflow")
experiment = client.create_experiment(name="ml-experiments")

# 두 함수의 공통 인자: arguments, run_name, experiment_name, namespace,
#   pipeline_root, enable_caching, cache_key, service_account, experiment_id
run = client.create_run_from_pipeline_package(
    pipeline_file="ml_pipeline.yaml",
    arguments={"n_estimators": 200, "accuracy_threshold": 0.90},
    run_name="training-run-001",
    experiment_id=experiment.experiment_id,
    enable_caching=True,
)

# CI 에서는 여기서 블로킹해야 종료 코드로 성패가 전달된다
client.wait_for_run_completion(run.run_id, timeout=3600, sleep_duration=5)

# 카탈로그 등록과 버전 올리기
client.upload_pipeline("ml_pipeline.yaml", pipeline_name="ml-training-v2")
client.upload_pipeline_version("ml_pipeline.yaml", "v3", pipeline_name="ml-training-v2")

# 정기 실행은 함수가 아니라 패키지나 등록된 파이프라인을 가리킨다
client.create_recurring_run(
    experiment_id=experiment.experiment_id,
    job_name="daily-retraining",
    pipeline_package_path="ml_pipeline.yaml",  # 또는 pipeline_id / version_id
    cron_expression="0 2 * * *",
    max_concurrency=1,
    no_catchup=True,      # 멈췄다 살아나도 밀린 스케줄을 몰아 돌리지 않는다
    params={"accuracy_threshold": 0.85},
)

from typing import NamedTuple

from kfp import compiler, dsl
from kfp.client import Client


@dsl.component(base_image="python:3.11")
def make_split(n_rows: int, ratio: float) -> NamedTuple('outputs', train=int, test=int):
    from typing import NamedTuple

    outputs = NamedTuple('outputs', train=int, test=int)
    n_test = int(n_rows * ratio)
    return outputs(n_rows - n_test, n_test)


@dsl.component(base_image="python:3.11", packages_to_install=["scikit-learn==1.4.0"])
def fit_and_score(
    train_rows: int,
    model: dsl.Output[dsl.Model],
    metrics: dsl.Output[dsl.Metrics],
):
    import json

    from sklearn.dummy import DummyClassifier

    X = [[0], [1]] * train_rows
    y = [0, 1] * train_rows

    clf = DummyClassifier(strategy="most_frequent")
    clf.fit(X, y)
    accuracy = float(clf.score(X, y))

    with open(model.path, "w") as f:
        json.dump({"strategy": "most_frequent"}, f)
    model.metadata["framework"] = "sklearn"
    metrics.log_metric("accuracy", accuracy)


@dsl.pipeline(name="smoke-pipeline", description="compile to submit smoke test")
def smoke_pipeline(n_rows: int = 1000, ratio: float = 0.2):
    split = make_split(n_rows=n_rows, ratio=ratio)

    fit = fit_and_score(train_rows=split.outputs["train"])
    fit.set_memory_request("512Mi")
    fit.set_memory_limit("1Gi")
    fit.set_retry(num_retries=2)


compiler.Compiler().compile(smoke_pipeline, package_path="smoke_pipeline.yaml")

client = Client(host="http://localhost:8080")
run = client.create_run_from_pipeline_package(
    "smoke_pipeline.yaml",
    arguments={"n_rows": 1000},
    run_name="smoke-001",
    experiment_name="smoke",
)
print(run.run_id)
client.wait_for_run_completion(run.run_id, timeout=900)

# 파라미터 매핑: str -> string, int/float -> number, bool -> boolean,
#               list/dict -> object
# 아티팩트 타입: dsl.Artifact(system.Artifact), Dataset, Model, Metrics,
#   ClassificationMetrics, SlicedClassificationMetrics, HTML, Markdown
#   공통 속성 .name .uri .path .metadata / Model 은 .framework 가 추가된다
# Metrics.log_metric(metric, value)
# ClassificationMetrics.log_roc_data_point(fpr, tpr, threshold), log_roc_curve(),
#   set_confusion_matrix_categories(), log_confusion_matrix_row(),
#   log_confusion_matrix_cell(), log_confusion_matrix(categories, matrix)
# dsl.InputPath / dsl.OutputPath 는 주로 Container Components 용이고,
#   Python Components 는 반환 타입 어노테이션으로 같은 일을 한다

1) 컴파일 산출물
   smoke_pipeline.yaml          <- 파일 하나뿐이다

2) 제출 직후 표준 출력
   런 ID가 UUID 한 줄로 찍힌다

3) UI 런 상세 화면의 그래프
   make_split       Succeeded
   fit_and_score    Succeeded
     Metrics  탭 : accuracy = 0.5
     Artifacts 탭 : model (system.Model), metadata.framework = sklearn

4) 같은 인자로 한 번 더 제출하면
   make_split       Succeeded  (캐시 적중 표시)
   fit_and_score    Succeeded  (캐시 적중 표시)
   -> 실행 시간이 몇 초로 줄어든다

# 동작하지 않는다 - 본문 바깥 심볼을 참조한다
import pandas as pd

TARGET_COLUMN = "label"


@dsl.component(base_image="python:3.11")
def bad_component(data: dsl.Input[dsl.Dataset]) -> int:
    df = pd.read_csv(data.path)                # NameError: name 'pd' is not defined
    return int(df[TARGET_COLUMN].nunique())    # NameError: name 'TARGET_COLUMN' ...


# 동작한다 - 모든 심볼이 본문 안에 있다
@dsl.component(base_image="python:3.11", packages_to_install=["pandas==2.1.4"])
def good_component(data: dsl.Input[dsl.Dataset], target_column: str = "label") -> int:
    import pandas as pd

    df = pd.read_csv(data.path)
    return int(df[target_column].nunique())

# 이미지만 만들고 푸시하지 않는다 (로컬 확인용)
kfp component build src/ --component-filepattern my_component.py --no-push-image

# 레지스트리까지 올린다 (CI에서 쓰는 형태)
kfp component build src/ --component-filepattern my_component.py --push-image

@dsl.component(base_image="python:3.11")
def notify(status: dsl.PipelineTaskFinalStatus):
    print("state:", status.state)
    if status.state == "FAILED":
        print("pipeline failed - send alert here")


@dsl.pipeline(name="pipeline-with-exit-handler")
def pipeline_with_exit_handler(n_rows: int = 1000):
    with dsl.ExitHandler(exit_task=notify()):
        split = make_split(n_rows=n_rows, ratio=0.2)

        fit = fit_and_score(train_rows=split.outputs["train"])
        fit.set_retry(num_retries=2)

# 1) 태스크 단위 - 이 태스크만 항상 새로 실행한다
load_task = load_data(dataset_url=dataset_url)
load_task.set_caching_options(False)

# 2) 런 단위 - 태스크 단위 설정을 덮어쓴다
run = client.create_run_from_pipeline_func(
    ml_training_pipeline,
    arguments={"dataset_url": "gs://my-bucket/data.csv"},
    enable_caching=False,
)

# 컴파일 플래그로 기본값을 끈다
kfp dsl compile --py my_pipeline.py --output my_pipeline.yaml \
  --disable-execution-caching-by-default

# 또는 환경변수로 (컴포넌트를 import 하기 전에 설정해야 한다)
export KFP_DISABLE_EXECUTION_CACHING_BY_DEFAULT=true
python my_pipeline.py

@dsl.pipeline(name="control-flow-example")
def control_flow_example(threshold: float = 0.85):
    # train_with_epochs, max_accuracy, promote_model, stage_model,
    # report_failure 는 각자 정의한 컴포넌트라고 가정한다

    # parallelism 으로 동시에 뜨는 Pod 수를 제한한다
    with dsl.ParallelFor(items=[1, 5, 10, 25], parallelism=2) as epochs:
        train_task = train_with_epochs(epochs=epochs)

    # 팬아웃한 결과를 하나로 모은다
    best = max_accuracy(models=dsl.Collected(train_task.outputs["model"]))

    with dsl.If(best.output >= threshold):
        promote_model(score=best.output)
    with dsl.Elif(best.output >= 0.70):
        stage_model(score=best.output)
    with dsl.Else():
        report_failure(score=best.output)

# PipelineTask 의 확인된 메서드 (전부 체이닝된다)
#   set_cpu_request, set_cpu_limit, set_memory_request, set_memory_limit,
#   set_accelerator_type, set_accelerator_limit, set_caching_options,
#   set_retry, set_env_variable, ignore_upstream_failure, after
#   set_gpu_limit 은 현재 API 레퍼런스에 없다

from kfp import dsl, kubernetes


@dsl.pipeline(name="pvc-example")
def pvc_example():
    pvc1 = kubernetes.CreatePVC(
        pvc_name_suffix='-my-pvc',
        access_modes=['ReadWriteMany'],
        size='5Gi',
        storage_class_name='standard',
    )

    task1 = producer()
    kubernetes.mount_pvc(task1, pvc_name=pvc1.outputs['name'], mount_path='/data')

    task2 = consumer().after(task1)
    kubernetes.mount_pvc(task2, pvc_name=pvc1.outputs['name'], mount_path='/data')

    # 정리까지 파이프라인 안에서 끝낸다
    kubernetes.DeletePVC(pvc_name=pvc1.outputs['name']).after(task2)


# 같은 패키지에서 확인된 다른 기능
#   use_secret_as_env, use_secret_as_volume, use_config_map_as_env,
#   use_config_map_as_volume, add_ephemeral_volume, add_pod_label,
#   add_pod_annotation, use_field_path_as_env, set_timeout,
#   set_image_pull_policy, set_security_context, set_image_pull_secrets