Focus on your models, not infrastructure. Simpletorch makes deploying PyTorch jobs on cloud GPUs effortless.
import torch
import simpletorch
@simpletorch.train(instance_type="g6.xlarge")
def train_model(ctx: simpletorch.TrainingContext, num_epochs: int = 25, lr: float = .001):
for epoch in range(num_epochs):
train_epoch(ctx.model, lr)
torch.save(model.state_dict(), "best.ckpt")
ctx.save("best.ckpt")
# Run normally - executes in the cloud!
if __name__ == "__main__":
model = train_model(my_classifier, "my_dataset/", "my_dependencies.txt", lr = .002)
Everything you need to train PyTorch models
Powerful GPUs
Access to a variety of GPUs for perfect training.
No Setup Time
Get started in minutes with pre-configured PyTorch environments. No DevOps required.
Easy Data Transfer
Simple upload interface for your datasets and models with high-speed cloud storage.
Real-time Monitoring
Logs stream straight to your console and are accessible through the web view.
Support
Round-the-clock support directly from the founder.
Full PyTorch Support
Latest PyTorch versions with all dependencies pre-installed.
Simple deployment in 3 steps
Create your PyTorch training function with your model, data loading, and training loop, just like normal.
Add the @simpletorch.train() decorator with your job config (GPU type, instance count, etc).
Run your Python script normally and it automatically executes in the cloud with full GPU access.
Simple pricing, no surprises
One flat rate gives you everything you need to train PyTorch models at scale.