https://github.com/zhreshold/gluoncv-sagemaker-serving
SageMaker serving for GluonCV models
https://github.com/zhreshold/gluoncv-sagemaker-serving
Last synced: about 1 year ago
JSON representation
SageMaker serving for GluonCV models
- Host: GitHub
- URL: https://github.com/zhreshold/gluoncv-sagemaker-serving
- Owner: zhreshold
- Created: 2020-04-29T21:24:08.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2020-08-21T23:43:15.000Z (almost 6 years ago)
- Last Synced: 2025-01-24T10:47:57.827Z (over 1 year ago)
- Language: Python
- Size: 235 KB
- Stars: 1
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# gluoncv-sagemaker
SageMaker serving for GluonCV models
# Pre-requisite
- SageMaker notebook(recommended)
- Computer with AWS CLI configured.
# Usage:
The following instructions works for `SageMaker` notebook instance(make sure ECR write permission is added in SageMaker IAM role)
- First activate a `mxnet_p36` env by `source activate mxnet_p36` in a new terminal.
- Then install required python packages: `pip install -r requirements.txt`
- Execute the batch building script: `python batch_build.py`.
The script will build `ModelPackage` for models listed in `image_classification.txt` and `object_detection.txt`. Due to the validation process, it might take very long time to finish.
# Marketplace listing Metadata Generator
```bash
python batch_describe.py
```
The sample output looks like: https://gist.github.com/zhreshold/f616f5b894d386701b4c85a4b40d200c
## TODO
- [ ] Automatic ARN injection for built ModelPackage
- [ ] Object Detection metadata
- [ ] Semantic Segmentation metadata