aws / aws/sagemaker-python-sdk

ModelStep should have a 'adds_depends_on' class method

已关闭
#5,440 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
component: pipelines type: bug
主要语言
Python
星标
2.3k
派生
1.3k
平均合并
1 天 22 小时
30 天内合并 PR
35

描述

**Describe the feature you'd like**
When constructing SageMaker pipelines, I want to construct steps that sequence Training jobs (TrainingStep) and Inference jobs (ModelStep and TransformStep) together. As part of this, I would use `add_depends_on` from the TrainingStep to allow me to string together tasks. However, we cannot do the same for ModelStep. It has to be passed via the `depends_on` parameter during initialization.

**How would this feature be used? Please describe.**
```
def setup(…):
model_step = ModelStep(…)
transform_step = TransformStep(…, model_step…)

model_step.add_depends_on(other_steps)
```

**Describe alternatives you've considered**
I am forced to pass in dependencies throughout my abstractions, rather than construct the model_step first, and then add dependencies later.

**Additional context**
happy to add more details if needed.

贡献指南

打开贡献指南

调研方向

首先定位 issue 中提到的 ModelStep、TrainingStep 和 TransformStep 入口点,然后比较 TrainingStep 如何处理 add_depends_on 与 ModelStep 的 depends_on 初始化参数。通过构建所示的 pipeline 序列并检查生成的依赖关系是否得到保留,确认所请求的用法。

由索引模型根据 Issue 内容生成。

评估

技术栈
aws, python
领域
machine-learning
Issue 类型
功能
难度
3/5
预计耗时
1-2 天
活跃度
停滞
描述清晰度
基本清楚
新手友好度
25/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。