lambda-labs-client 0.3.1

A Python client for Lambda Labs Cloud API for managing GPU instances

Lambda Labs API Client

A Python client and command-line tool for managing GPU instances on Lambda Labs Cloud. This package provides an easy way to list, launch, and monitor GPU instances with automatic retry capabilities for high-demand resources.

Features

• 📋 List running instances - View all your active instances with details
• 🖥️ List instance types - See all available GPU types (including unavailable ones)
• 🚀 Launch instances - Start new GPU instances with a single command
• 🔄 Automatic retry - Keep trying to launch instances until they become available
• 🔍 Smart filtering - Filter instance types by GPU model or availability status
• 💰 Cost tracking - See hourly pricing for all instance types
• 🛑 Terminate instances - Stop instances individually or all at once

Installation

From PyPI (Recommended)

pip install lambda-labs-client

From Source

git clone https://github.com/radekosmulski/lambda_labs_api.git
cd lambda_labs_api
pip install .

Development Installation

git clone https://github.com/radekosmulski/lambda_labs_api.git
cd lambda_labs_api
pip install -e .[dev]

Authentication

You need a Lambda Labs API key to use this client. Get one from the Lambda Labs API keys page.

Setting up your API key

You can provide your API key in two ways:

Option 1: Environment Variable (Recommended)

export LAMBDA_API_KEY="your-api-key-here"

Add this to your ~/.bashrc or ~/.zshrc to make it permanent.

Option 2: Command Line Argument

python lambda_labs_client.py --api-key "your-api-key-here" --list-instances

Usage

Basic Commands

List Running Instances

# Using environment variable
lambda-labs --list-instances

# Using API key argument
lambda-labs --api-key "your-key" --list-instances

List All Instance Types

# Show all instance types (available and unavailable)
lambda-labs --list-types

# Show only available instances
lambda-labs --list-types --show available

# Show only unavailable instances
lambda-labs --list-types --show unavailable

# Filter by GPU type
lambda-labs --list-types --filter-type a100
lambda-labs --list-types --filter-type h100

Launch an Instance

# Basic launch (will fail if unavailable)
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key"

# Launch with custom name
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key" --name "ML-Training"

# Launch in specific region
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key" --region us-east-1

# Launch multiple instances
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key" --quantity 2

Auto-Retry Feature

The killer feature: automatically retry launching instances until they become available!

# Keep trying every 5 seconds until successful
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key" --wait

# Retry with custom interval (every 10 seconds)
lambda-labs --launch gpu_8x_h100 --ssh-key "my-ssh-key" --wait --retry-interval 10

# Retry with maximum attempts
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key" --wait --max-retries 100

# Retry for specific region
lambda-labs --launch gpu_1x_a100 --ssh-key "my-ssh-key" --region us-west-1 --wait

Press Ctrl+C anytime to cancel the retry loop.

Terminating Instances

# Terminate specific instances by ID
lambda-labs --terminate 0920582c7ff041399e34823a0be62549

# Terminate multiple instances
lambda-labs --terminate instance-id-1 instance-id-2 instance-id-3

# Terminate ALL running instances
lambda-labs --terminate all

# Skip confirmation prompt (use with caution!)
lambda-labs --terminate all --force

Common Workflows

1. Check What's Available and Launch

# First, see what's available
lambda-labs --list-types --show available

# Then launch an available instance
lambda-labs --launch gpu_1x_a100 --ssh-key "my-key"

2. Wait for High-Demand GPUs

# Set up a retry for an H100 instance
lambda-labs --launch gpu_8x_h100 --ssh-key "my-key" --wait --retry-interval 30

3. Monitor Your Instances

# Check your running instances
lambda-labs --list-instances

4. Launch in Preferred Region

# Try to launch in us-east-1, but fall back to any available region
lambda-labs --launch gpu_1x_a100 --ssh-key "my-key" --region us-east-1 --wait

SSH Key Management

Before launching instances, you need to have SSH keys configured in your Lambda Labs account.

To see your available SSH keys:

lambda-labs --launch gpu_1x_a100
# (without specifying --ssh-key, it will list available keys)

Python API Usage

You can also use the client programmatically:

from lambda_labs_client import LambdaLabsClient

# Initialize client
client = LambdaLabsClient(api_key="your-api-key")

# List instances
instances = client.list_instances()
print(f"Found {len(instances)} running instances")

# Get instance types
instance_types = client.get_instance_types()
print(f"Available instance types: {len(instance_types)}")

# Launch an instance
instance_ids = client.launch_instance(
    region_name="us-west-1",
    instance_type_name="gpu_1x_a100",
    ssh_key_names=["my-ssh-key"],
    name="My Instance"
)
print(f"Launched instances: {instance_ids}")

# Terminate a specific instance
terminated = client.terminate_instance("instance-id-here")

# Terminate multiple instances
terminated = client.terminate_instances(["id1", "id2", "id3"])

# Terminate all running instances
terminated = client.terminate_all_instances()
print(f"Terminated {len(terminated)} instances")

Instance Type Examples

Common instance types you can launch:

• gpu_1x_a10 - 1x A10 (24 GB)
• gpu_1x_a100 - 1x A100 (40 GB)
• gpu_8x_a100 - 8x A100 (40 GB)
• gpu_1x_h100 - 1x H100 (80 GB)
• gpu_8x_h100 - 8x H100 (80 GB)

Output Examples

Listing Instances

================================================================================
                               RUNNING INSTANCES                                
================================================================================

Instance: ML-Training (ID: 0920582c7ff041399e34823a0be62549)
  Status: active
  Type: 1x A100 (40 GB SXM4)
  Region: California, USA (us-west-1)
  Public IP: 198.51.100.2
  Private IP: 10.0.2.100
  SSH Keys: my-ssh-key
  Jupyter URL: https://jupyter-url.lambdaspaces.com/?token=...
  Cost: $1.29/hour

Auto-Retry Output

Starting launch attempts for 'gpu_1x_a100'...
Retry interval: 5 seconds
Max retries: unlimited
Press Ctrl+C to cancel

[14:23:45] Attempt #1 (elapsed: 0s)
  ❌ No availability in any region
  ⏳ Waiting 5 seconds before next attempt...

[14:23:50] Attempt #2 (elapsed: 5s)
  ❌ No availability in any region
  ⏳ Waiting 5 seconds before next attempt...

[14:23:55] Attempt #3 (elapsed: 10s)
  ✅ Instance available in region 'us-west-1'! Launching...

🎉 Successfully launched 1 instance(s):
  - 0920582c7ff041399e34823a0be62549

Total wait time: 0m 10s

Tips

1. High-Demand GPUs: H100 and A100 instances are often fully booked. Use --wait to automatically grab one when it becomes available.

2. Overnight Launches: Set up a retry before going to bed:

   lambda-labs --launch gpu_8x_h100 --ssh-key "my-key" --wait --retry-interval 60
   

3. Cost Awareness: Always check the hourly cost shown in --list-types before launching.

4. Region Selection: Some regions have better availability. If you don't need a specific region, don't specify --region to allow the tool to pick any available region.

Error Handling

The client provides clear error messages:

• Invalid API Key: Check your API key or create a new one
• No SSH Keys: Add SSH keys to your Lambda Labs account first
• Instance Type Not Found: Use --list-types to see valid instance names
• No Availability: Use --wait to keep retrying

License

[Your chosen license]

Contributing

[Your contribution guidelines]

Support

For issues with the Lambda Labs API itself, contact Lambda Labs support. For issues with this client, please open a GitHub issue.