ml-pipeline-creation

一行命令安装

下载这个 Skill 需要先登录（免费注册，只需用户名和密码）。去登录 →

使用反馈 · 0

用过的同学可以在这里说说效果、坑点或改进建议，也可以直接提出新需求。

规范度体检 · 82/100

按元信息、示例、可执行性、完整性等 13 项自动检查，供作者自查改进；不影响用户评分。

• ✓ front-matter 含 name / description 15/15
• ✗ description 信息量（≥30 字） 2/5
  当前只有 24 字，建议写清做什么、解决什么问题
• ✓ description 含使用场景语境 5/5
• ✗ version 字段 0/3
  没有版本号，后续更新无法比对
• ✓ license 字段 2/2
• ✓ 正文篇幅（≥300 字） 10/10
• ✓ 含代码块 / 命令示例 10/10
• ✓ 结构化步骤 / 清单（≥3 条） 8/8
• ✗ 有边界 / 限制说明 0/6
  没写什么时候不该用，容易被误触发
• ✗ 有输出格式 / 验收标准 0/6
  缺少结果标准，Agent 输出质量不稳定
• ✓ 引用的文件均存在 18/18
• ✓ SKILL.md 体积合理 6/6
• ✓ 整体体积可控（≤2MB） 6/6

SKILL.md 内容

## Workflow

This skill enables the creation and management of machine learning (ML) pipelines, automating the process of training, evaluating, and deploying ML models. The workflow is designed to be flexible and adaptable to various ML tasks and frameworks.

1.  **Define Pipeline Structure**: The user specifies the stages of the ML pipeline, including data preprocessing, model training, model evaluation, and deployment. This is typically done in a configuration file (e.g., YAML or JSON).
2.  **Component Implementation**: Each stage of the pipeline is implemented as a separate component. These components are reusable and can be chained together to form a complete pipeline.
3.  **Pipeline Execution**: The skill executes the pipeline, running each component in the specified order. It handles data flow between components and manages dependencies.
4.  **Monitoring and Logging**: The skill provides tools for monitoring the pipeline's execution, logging results, and tracking experiments.
5.  **Deployment**: Once a model is trained and evaluated, the skill can automate its deployment to a serving environment.

## Usage

To use this skill, you need to provide a pipeline definition file and the implementation of the pipeline components.

### Example: Simple Scikit-learn Pipeline

Here's an example of how to define and run a simple ML pipeline using this skill.

**`pipeline.yaml`**

```yaml
name: simple-sklearn-pipeline
components:
  - name: data-preprocessing
    script: preprocess.py
    inputs:
      - raw_data: /path/to/raw_data.csv
    outputs:
      - processed_data: /path/to/processed_data.csv
  - name: train-model
    script: train.py
    inputs:
      - processed_data: /path/to/processed_data.csv
    outputs:
      - model: /path/to/model.pkl
  - name: evaluate-model
    script: evaluate.py
    inputs:
      - model: /path/to/model.pkl
      - test_data: /path/to/test_data.csv
    outputs:
      - metrics: /path/to/metrics.json
```

**`preprocess.py`**

```python
import pandas as pd
from sklearn.model_selection import train_test_split

# Load data
df = pd.read_csv('/path/to/raw_data.csv')

# Simple preprocessing
X = df.drop('target', axis=1)
y = df['target']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Save processed data
pd.concat([X_train, y_train], axis=1).to_csv('/path/to/processed_data.csv', index=False)
pd.concat([X_test, y_test], axis=1).to_csv('/path/to/test_data.csv', index=False)
```

**`train.py`**

```python
import pandas as pd
from sklearn.ensemble import RandomForestClassifier
import joblib

# Load processed data
df = pd.read_csv('/path/to/processed_data.csv')
X_train = df.drop('target', axis=1)
y_train = df['target']

# Train model
model = RandomForestClassifier(n_estimators=100, random_state=42)
model.fit(X_train, y_train)

# Save model
joblib.dump(model, '/path/to/model.pkl')
```

**`evaluate.py`**

```python
import pandas as pd
import joblib
import json
from sklearn.metrics import accuracy_score

# Load model and test data
model = joblib.load('/path/to/model.pkl')
df = pd.read_csv('/path/to/test_data.csv')
X_test = df.drop('target', axis=1)
y_test = df['target']

# Evaluate model
y_pred = model.predict(X_test)
accuracy = accuracy_score(y_test, y_pred)

# Save metrics
with open('/path/to/metrics.json', 'w') as f:
    json.dump({'accuracy': accuracy}, f)

print(f'Model accuracy: {accuracy}')
```