Skip to content
No description, website, or topics provided.
Python Shell
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

ds run demo

NOTE: This demo requires a runner with a GPU

To run this using ds run, try the following:


Make a new project in Dotscience:

ds project create circleci

Create a runner

ds runner create gpu-runner-x -s 100 -t gpu-nvidia-runtime

It should then display setup instructions (just with your own details and Dotscience environment):

Runner ID:        207142bc-da9f-4ca1-a58e-f06a3c0ad998
Runner API token: [YOUR TOKEN]

To start a runner, run:

docker pull && \
docker run --name dotscience-runner -d -e TOKEN=[YOUR TOKEN] \
--restart always -v /var/run/docker.sock:/var/run/docker.sock \
-v dotscience-task-spool:/spool \ ds-runner run --addr 

You can copy/paste the output from ds runner create command into a VM that has GPU and Docker installed.

To view available runners, use ds runner ls command:

$ ds runner ls
ID                                     NAME                RUNNING TASKS       STATUS              TYPE                 AGE
207142bc-da9f-4ca1-a58e-f06a3c0ad998   gpu-runner-x        0                   offline             gpu-nvidia-runtime   13 minutes
c6d6f46f-4a32-4a9d-b649-acc28b282bc3   worker-x            0                   online              CPU                  About an hour

Setting up GitHub authentication to clone repository

Dotscience can use SSH based authentication which is compatible with all source control management solutions. For the sake of simplicity, we will use GitHub.

To generate a new private/public SSH key pair, use ds secret generate --name [NAME] command:

$ ds secret generate --name deployment-key
Public key: 
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQC67QalGM8aAcynRNqtizyn8ZKDuVE7cue63DRPfdo+NdNArZvAhp0wS+yREpomh6XtDhTXhgZeuzJLbfBPPsHBmLx+kDR0TQh8Y5ISoEjIGQHOuUPwlsIrD2JwVI4AheCikKIVIJU4UIuvZgNErJo/3zeCkMVDgMVrGrCDQVh/Eanxm9VKid5YFY3no4j88Nf3KwCtK0fMo93xNDS35RvezjuKCmxg91hnleDBozKNpck19uPk5Ww957QmwQYNEWKVB3xoa/SKUSxSksJJizhuos1vrooG7EX8b3JoIKyD/i92pbSM6mMeF0FmBuRWEU4EMsgbyFtp4S44u6utxaNF

Now, you can copy/paste this key into your Github repository settings under<username or org name>/<repo name>/settings/keys.

Once it's added, Dotscience will be able to clone this repository. You can add multiple keys and remove them if they are not needed anymore. Keys can be attached to a specific project or for any project in the account (if you don't set project ID when creating a secret).

Fetching data

Every ds run command will need a project name. If you forgot your project name, type ds project ls:

$ ds project ls
ID                                     NAME                DOTS                                   COLLABORATORS       AGE
ac6d7708-6988-4b49-b922-f3be7bdeaaf7   circleci            8c35ec84-9382-4263-a7e4-2f98885540e6                       3 hours
ds run --project-name circleci --nvidia --image --repo --ref master bash ds-run-demo/

Note that the actual command can be specified either with -c flags or just positional arguments bash ds-run-demo/ You will need to specify a repository name as Dotscience clones repository into a subdirectory named after the repository. Any changes to this repository directory will be overwritten during next checkout.

Train model

Now, to test train on master branch:

ds run --project-name run-demo --nvidia --image --repo --ref master python ds-run-demo/

You can replace flag --ref value with a commit ID, other branch name or tag name to train on that specific point in time.

Check score

To see model performance, use ds runs ls [PROJECT NAME OR ID] command:

ds runs ls run-demo
LABELS                                                                                     PARAMETERS                  TYPE                SCORE               AGE
model.framework=tensorflow, model.framework.version=1.13.0-rc0,      epochs=1, optimizer=adam    accuracy            0.9030879           24 hours
model.framework=tensorflow, model.framework.version=1.13.0-rc0,   epochs=10, optimizer=adam   accuracy            0.9332542           24 hours, model.framework=tensorflow, model.framework.version=1.13.0-rc0   epochs=5, optimizer=adam    accuracy            0.9219319           5 hours, model.framework=tensorflow, model.framework.version=1.13.0-rc0   epochs=5, optimizer=adam    accuracy            0.919715            37 minutes, model.framework=tensorflow, model.framework.version=1.13.0-rc0   epochs=5, optimizer=adam    accuracy            0.9275534           20 minutes

You should see provenance & metrics data in Dotscience!

You can’t perform that action at this time.