Create A Projectand give your project a name.
Workflowstab we'll now select
Createto initialize our first Workflow.
Get started with workflowspage.
Repository namewe have inputted "Gradient Workflows Tutorial" -- this will be the name of the repository that Gradient writes to GitHub on our behalf.
Select account or organization.
Repository namefor writing to GitHub. In our case we will name our repository "Gradient Workflows Tutorial."
Create Workflowbutton after which we should see the new workflow begin to initialize:
Succeededin the left sidebar. To confirm this, we can see that the boxes containing each of our jobs (
StyleGan2) now display the checkmark symbol.
StyleGan2, we should see something like this:
Logstab for the moment and view the
Datatab within the
StyleGAN2job. Here we'll find the seed images that we have just generated.
Dataview for the
StyleGAN2job and inspect the seed files that we've generated.
seed6600.pngshould look exactly like the image above. Reproducibility for the win!
Gradient-Workflows-Tutorialpreviously so our git clone command will look like this. Yours will be different depending on the GitHub account you connected to Gradient and the name you gave to your first Workflow.
cdinto the local directory and
lsto reveal a directory structure like this:
ls -acommand, we will reveal a number of hidden directories including
.gradientdirectory and we can
cdinto it and
lsto confirm that it contains a
workflowsdirectory that contains our YAML file.
stylegan2.yamlwhich is available at
/.gradient/workflows/stylegan2.yamlwithin the local directory that we cloned from GitHub.
.gradient/workflowswithin the repo.
CloneRepoand the other is called
StyleGAN2. This is exactly what we expected to see since we can recall these two jobs running in Part 1.
mainbranch we will trigger a new workflow run.
.gradient/workflows/stylegan2.yamlwith this new YAML file.
onblock is commented out. We will need to uncomment this block before we push the file to GitHub if we're going to run the new workflow.
jobsheading in the file.
cloneStyleGAN2Repo- uses a C4 instance type (a basic CPU machine) to clone the StyleGAN2 source code from NVIDIA's GitHub repo into a managed storage provider on Gradient
getPretrainedModelCats- uses a C4 instance type to copy a pretrained model of cats (a large
.pklfile) over to Gradient
generateCatImagesPretrainedModel- uses a P4000 instance type to display images of cats generated using the pretrained cats model
getPretrainedModelCars- uses a C4 instance type to copy a pretrained model of cars (a large
.pklfile) over to Gradient
generateCarImagesPretrainedModel- uses a P4000 instance type to display images of cars generated using the pretrained cars model
onblock our workflow will not run.
modify yaml file) and then we are seeing the commit has triggered a new workflow run.
generateCatImagesPretrainedModel-- our expectation is to find images of cats within the dataset.
DATAtab to explore the dataset that we've just generated. Inside we should find a number of seed images of cats.
uncomment steps 4 and 5.
uncomment steps 4 and 5