A CLI tool to manage your models.
It takes a TensorFlow SaveModel file, uploads it with version and meta info, and checkouts it to local.
- info model_name[@version_number]
- create model_name -m “comment”
- submit model_name -m “comment” [–batching_config=batching_config.txt] model_path
- checkout model_name[@version_number]
- deploy model_name[@version_number] [CPU/GPU] [online/offline] [instance-number] [offline]
## NTF models
ModelHub ogirinally supports uploading only TensorFlow-based models. From 0.0.8, it supports NTF models.
To prepare NTF models, you need create a new folder containing two subfolders: variables and assets
- varibles: storing trained models
- assets: storing any other data files, for example, feature dictionaries.
## Python API pass get model path by model name [and version]