conduct lambda

GPU cluster management for Lambda Labs - system specs, monitoring, visualizations, and deployment workflows.

Command Hierarchy

lambda — GPU cluster management │ ├── specs System specifications ├── models LLM model catalog │ ├── monitor Real-time GPU dashboard │ -r, --refresh Refresh interval (seconds) │ -o, --once Single snapshot mode │ ├── visualize GPU visualizations │ ├── gpu Memory layout & KV cache │ ├── heat Thermal topology heat map │ └── usage Resource utilization bars │ ├── workflows Deployment presets │ ├── list Browse available workflows │ ├── show Workflow details & diagram │ │ -d, --diagram ASCII architecture diagram │ │ -m, --memory-limit Memory cap percentage │ ├── run Execute workflow │ │ -n, --dry-run Preview without executing │ └── coverage Coverage analysis presets │ ├── install Install vLLM & models ├── optimize Inference optimizations ├── config Server configuration │ ├── packages System package management │ └── tree This command hierarchy
root level 1 level 2 flags

Synopsis

conduct lambda [subcommand] [flags]

Quick Start

# Check system & GPU info
conduct lambda specs

# Real-time GPU dashboard
conduct lambda monitor

# Browse deployment presets
conduct lambda workflows list

# Show command hierarchy
conduct lambda tree

Subcommands

Commands Primary

Command Description
specs Display comprehensive system specifications (CPU, GPU, memory, disk)
models Browse available LLM catalog with specs and deployment info
tree Display command hierarchy tree with depth-based coloring

Monitoring Monitor

Command Description
monitor Live GPU metrics dashboard with inference stats, queue status, error rates
visualize gpu GPU memory layout with color-coded model blocks & KV cache
visualize heat Thermal topology heat map across all GPUs
visualize usage Resource utilization bars (CPU, GPU, memory, disk I/O)

Workflows Workflow

Command Description
workflows list Browse all available deployment workflow presets
workflows show <name> View workflow details with architecture diagram
workflows run <name> Execute workflow on current instance
workflows coverage Coverage analysis workflow presets

Setup Setup

Command Description
install Install vLLM and download models (passthrough to lambda CLI)
optimize Apply inference optimizations (Medusa, lookahead, batch)
config Server configuration management (save, load, show, edit)

Advanced Advanced

Command Description
packages System package management (list, suggest, install, check)

Flags

monitor

Flag Type Default Description
-r, --refresh int 2 Refresh interval in seconds
-o, --once bool false Single snapshot, exit immediately

models

Flag Type Default Description
-j, --json bool false Output as JSON for API consumption
-f, --filter string Filter models by name or size (e.g., "70B", "llama")
-w, --workflows bool false Show available deployment workflows

workflows show

Flag Type Default Description
-d, --diagram bool false Show ASCII GPU allocation diagram
-m, --memory-limit int 90 Memory usage limit percentage

workflows run

Flag Type Default Description
-n, --dry-run bool false Preview without executing
-m, --memory-limit int 90 Memory usage limit percentage
-v, --verbose bool false Show detailed output for each step

Examples

System Information

# View comprehensive system specs
conduct lambda specs

# Browse model catalog
conduct lambda models

# Filter models by size
conduct lambda models --filter 70B

# Export models as JSON
conduct lambda models --json

Monitoring

# Launch real-time GPU dashboard
conduct lambda monitor

# Single snapshot
conduct lambda monitor --once

# Custom refresh interval (5 seconds)
conduct lambda monitor --refresh 5

# GPU memory visualization
conduct lambda visualize gpu

# Thermal heat map
conduct lambda visualize heat

# Resource usage bars
conduct lambda visualize usage

Workflows

# List available workflows
conduct lambda workflows list

# View workflow details with diagram
conduct lambda workflows show dual-llama-70b-tp4 --diagram

# Preview workflow execution
conduct lambda workflows run dual-llama-70b-tp4 --dry-run

# Execute workflow
conduct lambda workflows run dual-llama-70b-tp4 --verbose

Setup & Configuration

# Install vLLM
conduct lambda install vllm

# Install recommended models
conduct lambda install models --recommended

# Apply all optimizations
conduct lambda optimize all

# View current configuration
conduct lambda config show

# Save configuration
conduct lambda config save my-config.yaml
Recommended Workflow

For new instances, follow this sequence:

  1. lambda specs - Verify hardware
  2. lambda workflows list - Choose deployment preset
  3. lambda workflows run <name> - Deploy models
  4. lambda monitor - Verify deployment health