Use cases
AI compute for training

Model Training GPU Workspaces

Spin up training workspaces for PyTorch experiments, dataset preparation, checkpoints, and repeatable GPU development.

Ready template

Recommended workspace

Template

PyTorch

GPU

H100

Access

Browser + SSH

Billing

Per-hour

porfal create --use-case model-trainingworkspace ready: browser + ssh

Workloads

PyTorch training

Experiment runs

Checkpointing

Data preprocessing

Recommended stack

PyTorch

JupyterLab

Ubuntu CUDA

TensorFlow

GPU fit

H100

H200

L40S

Why Porfal

Dedicated GPU compute without infrastructure work.

Pick a GPU, launch the template, connect from browser or SSH, and stop the workspace when the run is done.

Match GPU memory to model size.

Use templates instead of rebuilding environments.

Stop workspaces when training windows end.

Workflow

A short path from GPU to running workload.

1

Select training GPU

2

Choose PyTorch

3

Connect via SSH

4

Run training job