https://github.com/kwame-mintah/aws-lambda-model-training
A lambda function split preprocessed data into training and validation used for starting a training job within AWS SageMaker.
https://github.com/kwame-mintah/aws-lambda-model-training
aws aws-lambda model-training python python311 sagemaker
Last synced: 3 months ago
JSON representation
A lambda function split preprocessed data into training and validation used for starting a training job within AWS SageMaker.
- Host: GitHub
- URL: https://github.com/kwame-mintah/aws-lambda-model-training
- Owner: kwame-mintah
- Created: 2024-02-21T18:52:10.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-08-17T11:05:35.000Z (about 1 year ago)
- Last Synced: 2025-02-01T11:43:46.705Z (8 months ago)
- Topics: aws, aws-lambda, model-training, python, python311, sagemaker
- Language: Python
- Homepage:
- Size: 123 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
Awesome Lists containing this project
README
# AWS Lambda Model Training
[](https://www.python.org/downloads/release/python-3121/)
[](https://github.com/kwame-mintah/aws-lambda-model-training/actions/workflows/bump-repository-version.yml)
[](https://github.com/kwame-mintah/aws-lambda-model-training/actions/workflows/push-docker-image-to-aws-ecr.yml)
[](https://github.com/kwame-mintah/aws-lambda-model-training/actions/workflows/run-python-linter.yml)
A lambda to split pre-processed data into, training, validation and test datasets uploaded to an S3 bucket. Training and validation
datasets will be used when starting the AWS SageMaker training job and the test data will be used during [model evaluation](https://github.com/kwame-mintah/aws-lambda-model-evaluation).
This repository does not create the AWS resources, this is created via Terraform found here [terraform-aws-machine-learning-pipeline](https://github.com/kwame-mintah/terraform-aws-machine-learning-pipeline).
For more details on the entire flow and how this lambda is deployed, see [aws-automlops-serverless-deployment](https://github.com/kwame-mintah/aws-automlops-serverless-deployment).
# Flowchart
The [diagram below](https://mermaid.js.org/syntax/flowchart.html#flowcharts-basic-syntax) demonstrates what happens when the lambda is trigger, when a new `.csv` object has been uploaded to the S3 Bucket.
```mermaid
graph LR
S0(Start)
T1(Dataset pulled from S3 Bucket)
T2(Random split and sort using Numpy)
T3[["`70% training data
20% validation data
10% test data`"]]
T4("Upload split data into S3 Bucket as `.csv`")
T5("Start training job with training and validation data")
E0(End)
S0-->T1
T1-->T2
T2-->T3
T3-->T4
T4-->T5
T5-->E0
```
# Notice
The code provided here should serve as an example for creating a lambda function to start a AWS SageMaker training job.
Training algorithm used within this project is [XGBoost](https://docs.aws.amazon.com/sagemaker/latest/dg/xgboost.html) so
hyperparameters may not be suitable for all algorithms.
Additionally, the hyperparameters values are currently hardcoded, one approach could be to read from a `manifest.json`
file detailing what parameters and value to use for each dataset received. This would allow for more flexibility, so
a new docker image does not need to be deployed, when hyperparameters need to be changed.
## Development
### Dependencies
- [Python](https://www.python.org/downloads/release/python-3121/)
- [Docker for Desktop](https://www.docker.com/products/docker-desktop/)
- [Amazon Web Services](https://aws.amazon.com/?nc2=h_lg)
## Usage
1. Build the docker image locally:
```shell
docker build --no-cache -t model_training:local .
```
2. Run the docker image built:
```shell
docker run --platform linux/amd64 -p 9000:8080 model_training:local
```
3. Send an event to the lambda via curl:
```shell
curl "http://localhost:9000/2015-03-31/functions/function/invocations" -d '{}'
```
Example AWS S3 event received
```json
{
"Records": [
{
"eventVersion": "2.0",
"eventSource": "aws:s3",
"awsRegion": "us-east-1",
"eventTime": "1970-01-01T00:00:00.000Z",
"eventName": "ObjectCreated:Put",
"userIdentity": { "principalId": "EXAMPLE" },
"requestParameters": { "sourceIPAddress": "127.0.0.1" },
"responseElements": {
"x-amz-request-id": "EXAMPLE123456789",
"x-amz-id-2": "EXAMPLE123/5678abcdefghijklambdaisawesome/mnopqrstuvwxyzABCDEFGH"
},
"s3": {
"s3SchemaVersion": "1.0",
"configurationId": "testConfigRule",
"bucket": {
"name": "example-bucket",
"ownerIdentity": { "principalId": "EXAMPLE" },
"arn": "arn:aws:s3:::example-bucket"
},
"object": {
"key": "automl/example-bank-file.csv",
"size": 515246,
"eTag": "0e29c0d99c654bbe83c42097c97743ed",
"sequencer": "00656A54CA3D69362D"
}
}
}
]
}
```
## GitHub Action (CI/CD)
The GitHub Action "🚀 Push Docker image to AWS ECR" will check out the repository and push a docker image to the chosen AWS ECR using
[configure-aws-credentials](https://github.com/aws-actions/configure-aws-credentials/tree/v4.0.1/) action. The following repository secrets need to be set:
| Secret | Description |
|--------------------|------------------------------|
| AWS_REGION | The AWS Region. |
| AWS_ACCOUNT_ID | The AWS account ID. |
| AWS_ECR_REPOSITORY | The AWS ECR repository name. |