terraform-provider-iterative
terraform-provider-iterative copied to clipboard
iterative
☁️ Terraform plugin for machine learning workloads: spot instance recovery & auto-termination | AWS, GCP, Azure, Kubernetes
# Sagemaker It will: - Upload your training script and dependencies to Amazon S3 - Provision the specified number of instances in a fully managed cluster - Pull the specified...
It would be nice[^1] to have an `Execute` method to run commands in any of the active instances. https://github.com/iterative/terraform-provider-iterative/blob/57625b6db38238a18dd10797ac239710c5f94f8a/task/task.go#L37-L47 ### Loose ends * How to deal with parallelism? * Run...
Right now even after the task execution we are still maintaining the resources created. A task should be just only a resource linked to the bucket and not to all...
Should probably use https://pre-commit.ci with https://github.com/dnephin/pre-commit-golang or https://github.com/TekWizely/pre-commit-golang.
Unless we have any quota issues, the Kubernetes cluster should be created in the same region as the other test resources. https://github.com/iterative/terraform-provider-iterative/blob/352cd13cfa79837334ea1524d7d19071053900f7/.github/workflows/smoke.yml#L79 https://github.com/iterative/terraform-provider-iterative/blob/352cd13cfa79837334ea1524d7d19071053900f7/task/task_test.go#L59
We should show up when users search for "mlops": https://registry.terraform.io/search/providers?q=mlops
Goal: the interface for all our orchestrator tools should be similar including TFI, CML, TFI-Py and `dvc machine` Users should be able to move their projects from TFI to DVC...
Replaced the deprecated nvidia package with the new `nvidia-container-utils` as mentioned in the official [documentation](https://github.com/NVIDIA/nvidia-docker/tree/36e7a11c52f300668099ee500791e07302810007)
> With the release of Docker 19.03, usage of nvidia-docker2 packages is deprecated since NVIDIA GPUs are now natively supported as devices in the Docker runtime.
This closes #299. The storage type is inferred by provider. The `storage` config section is extended with additional options: ``` storage { container = container_path = container_opts = { }...
