Shadeform Quickstart Guide

Quickstart Guide

GPU instances with Shadeform's API

Export

Run Notebooks

idle

Contents

No cells yet

Add cells to see them here

🚀 Shadeform GPU Instances - Quickstart Guide

Welcome to the Shadeform API Quickstart Guide! This notebook will walk you through the complete lifecycle of GPU instances on Shadeform:

Step	Operation	Description
1️⃣	Find	Search for available GPU instances
2️⃣	Launch	Create and spin up your instance
3️⃣	Monitor	Check instance status until ready
4️⃣	Connect	SSH into your GPU instance
5️⃣	Manage	Restart or delete when done

💡 Tip: All operations can also be performed via the Shadeform UI

📋 Prerequisites

Before getting started, make sure you have completed these two steps:

Requirement	Action	Link
💰 Wallet Balance	Top up your account wallet	Add Funds →
🔑 API Key	Generate your Shadeform API key	Get API Key →

⚠️ Important: Keep your API key secure and never commit it to public repositories!

[1]

✅ Setup complete! Libraries imported and API configured.

1️⃣ Finding a GPU Instance

Use the /instances/types endpoint to search for available GPU instances. You can filter by:

Parameter	Description	Example
`gpu_type`	Type of GPU	`A6000`, `A100`, `H100`
`num_gpus`	Number of GPUs	`1`, `2`, `4`, `8`
`available`	Only show available instances	`true`
`sort`	Sort results	`price` (cheapest first)

Let's find the cheapest available A6000 instance:

[ ]

2️⃣ Launching the Instance

Now let's launch our GPU instance using the /instances/create endpoint.

Required Parameters:

Parameter	Description
`cloud`	Cloud provider (from search results)
`region`	Deployment region
`shade_instance_type`	Instance type identifier
`shade_cloud`	Use Shadeform's managed environment
`name`	A friendly name for your instance

⏱️ Note: Instance creation typically takes 3-5 minutes depending on the GPU type.

[2]

⚠️ Demo mode: Uncomment the launch_instance() call above to create a real instance

3️⃣ Checking Instance Status

Monitor your instance using the /instances/{id}/info endpoint. The instance will transition through these states:

	┌──────────┐     ┌──────────┐     ┌──────────┐
│ pending  │ ──► │  active  │ ──► │ deleted  │
└──────────┘     └──────────┘     └──────────┘
     │                                  ▲
     └──────────► (error) ──────────────┘

Status	Description
`pending`	Instance is being provisioned
`active`	Ready to use! IP address available
`deleted`	Instance has been terminated

[ ]

4️⃣ Connecting via SSH

Once your instance is active, you can SSH into it.

Step 1: Download Your Private Key

Download your Shadeform-generated private key from Settings → SSH Keys

Step 2: Set Correct Permissions

	chmod 600 ~/Downloads/private_key.pem

Step 3: Connect!

	ssh -i ~/Downloads/private_key.pem shadeform@<your-instance-ip>

🔧 Troubleshooting: If you see WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED!, run:
	ssh-keygen -R <your-instance-ip>

[3]

🔐 SSH CONNECTION DETAILS
==================================================

    📍 IP Address:    123.45.67.89
    👤 Username:      shadeform
    🔌 Port:          22
    🔑 Key File:      ~/Downloads/private_key.pem
    
    📋 Copy this command to connect:
    ┌──────────────────────────────────────────────────
    │  ssh -i ~/Downloads/private_key.pem -p 22 shadeform@123.45.67.89
    └──────────────────────────────────────────────────
    
    🔧 If you get a host key error, run:
    │  ssh-keygen -R 123.45.67.89

'ssh -i ~/Downloads/private_key.pem -p 22 shadeform@123.45.67.89'

5️⃣ Managing Your Instance

Restart Instance

Use the /instances/{id}/restart endpoint to reboot your instance.

⚠️ A 200 response confirms the restart was initiated, not completed. Wait a few minutes before reconnecting.

Delete Instance

Use the /instances/{id}/delete endpoint to terminate and remove your instance.

💡 Important: Always delete instances when you're done to avoid unnecessary charges!

[ ]

🛠️ Bonus: All-in-One Helper Class

Here's a convenient wrapper class that combines all the functionality above:

[4]

✅ ShadeformClient class is ready to use!

Example workflow:
─────────────────────────────────────────────
client = ShadeformClient("your-api-key")

# Find cheapest A6000
instances = client.find_instances(gpu_type="A6000")

# Launch it
instance_id = client.launch(instances[0], name="my-gpu")

# Wait for it to be ready
info = client.wait_until_active()

# Get SSH command
print(client.ssh_command())

# When done, delete it
client.delete()
─────────────────────────────────────────────

📚 API Reference Summary

Endpoint	Method	Description
`/instances/types`	GET	Search available instance types
`/instances/create`	POST	Launch a new instance
`/instances/{id}/info`	GET	Get instance details
`/instances/{id}/restart`	POST	Restart an instance
`/instances/{id}/delete`	POST	Terminate an instance

Available GPU Types

GPU	VRAM	Best For
A6000	48 GB	Training, fine-tuning medium models
A100	40/80 GB	Large model training, inference
H100	80 GB	Cutting-edge LLM training
RTX 4090	24 GB	Inference, smaller training jobs

🔗 Additional Resources

Happy GPU Computing! 🚀

Built with ❤️ using Shadeform