The remote client allows you to create model experiments using the command line instead of relying on API calls.
pip3 install -y nvidia-tao-client
Below is a detectnet_v2 example of shared storage content.
shared
└── users
├── 1a40ddb6-d389-4c39-9b40-eaa948ee508d
│ ├── metadata.json
│ ├── datasets
│ │ ├── 78cff931-e867-4e9c-b8a7-30fa243e433c
│ │ │ ├── metadata.json
│ │ │ ├── specs
│ │ │ │ └── convert.json
│ │ │ ├── images
│ │ │ │ ├── 000000.jpg
│ │ │ │ └── …
│ │ │ ├── labels
│ │ │ │ ├── 000000.txt
│ │ │ │ └── …
│ │ │ ├── 444fed7a-ac3f-4499-8a57-0234a43c752d
│ │ │ │ ├── status.json
│ │ │ │ └── …
│ │ │ └── tfrecords
│ │ │ └── …
│ │ └── …
│ └── models
│ ├── cdf387d2-067e-4935-be9d-7facbc851f60
│ │ ├── metadata.json
│ │ ├── specs
│ │ │ ├── train.json
│ │ │ └── …
│ │ ├── 444fed7a-ac3f-4499-8a57-0234a43c754e
│ │ │ ├── status.json
│ │ │ └── …
│ │ └── experiment_dir_unpruned
│ │ └── …
│ └── …
└── …
The command line depends on the following environment variables:
BASE_URL
USER
TOKEN
The command line is composed of:
$ tao-client <network> <action> <args>
For example:
$ tao-client detectnet-v2 dataset-convert --id 042559ec-ab3e-438d-9c94-2cab38f76efc
You can always use –help argument to retrieve the command usage information.
To list supported networks:
$ tao-client –help
To list supported detectnet-v2 actions:
$ tao-client detectnet-v2 –help
To list detectnet-v2 dataset-convert arguments:
$ tao-client detectnet-v2 dataset-convert --help
The tutorial notebooks provide examples that demonstrate the various workflows.
Download the resource using the NGC CLI.
ngc registry resource download-version "nvidia/tao/tao-getting-started:4.0.0"
Find tutorial notebooks (
.ipynbfiles).cd tao-getting-started_v4.0.0/notebooks/tao_api_starter_kit/client
Serve these Notebook examples using Jupyter.
pip3 install jupyter notebook jupyter notebook --ip 0.0.0.0