Quick Start Guide

This guide will help you get started with Serverless GPU API quickly.

Basic Usage

Here’s a simple example of how to use Serverless GPU API:

from serverless_gpu.launcher import distributed
from serverless_gpu.compute import GPUType

@distributed(gpus=8, gpu_type=GPUType.H100, remote=True)
def my_function(**kwargs):
    # Your code here
    pass

my_function.distributed(x=5)

For parameter sweeps, here’s another example:

from serverless_gpu import distributed

@distributed(
    gpus=2,
    gpu_type='a10',
    remote=True, # Uses remote gpus, not the gpus that are attached to your notebook
    run_async=True, # Enables multiple async workloads at once
)
def my_training_function(learning_rate: float):
  # TRAINING CODE HERE

# Launch multiple async executions
calls = []

for lr in [1e-3, 1e-4, 1e-5]
    # Launches the training code in my_training_function with different learning rates!
    calls.append(my_training_function.distributed(lr))

# Wait for completion
wait(calls)

# Your distributed computation code here
# Serverless GPU API will handle the multi-GPU orchestration

Next Steps

• Read the Overview to understand Serverless GPU API’s architecture

• Check out the API Reference for detailed API documentation

• Explore other examples in the repository