Skip to main content

Quickstart

quick-start

For Data Owners: If your environment is already set up, your client is running and you want to get benchmarks on your proprietary data, jump directly to create a use case.

Data Scientists: Join a use case and start training here.

For Data Owners​

First-Time Setup​

1. Environment Setup​

Set up your environment and compute cluster locally on a laptop or on large on-prem or cloud clusters. The process includes:

  • Reviewing hardware and software requirements
  • Setting up a Kubernetes cluster
  • Configuring necessary namespaces, resource quotas and security policies
  • Setting up storage for data and models

To get started, refer to the Environment Setup section.

2. Create Your Client​

Once your computation cluster is ready, you'll need to create and deploy a tracebloc client on your infrastructure. This involves:

  • Pulling the tracebloc client and registering on in the tracebloc platform
  • Generating and securing client credentials
  • Configuring Helm for deployment
  • Deploying the client using Helm charts

3. Create a Use Case​

Once the client is deployed, create a use case and prepare your data. This includes:

  • Setting up a data ingestion pipeline for training and test datasets
  • Inviting data scientists to your use case
  • Configuring access controls and permissions

For Data Scientists​

If you are a data scientist and want to prove the value of your solution, start here. This documentation includes how to connect to a client, run trainings, or fine-tune.

If you are a data scientist looking to demonstrate the performance of your models, start here. This section will guide you through:

  • Connecting to a tracebloc client
  • Running inference and training jobs
  • Fine-tuning and validating your models on client infrastructure