Skip to content

ui-types.md

Latest commit

 

History

History
132 lines (101 loc) · 4.04 KB

ui-types.md

File metadata and controls

132 lines (101 loc) · 4.04 KB
sidebar-title
User Interface

User Interface

AIPerf provides 3 UI types to display benchmark progress.

Overview

Choose the UI type that matches your workflow:

  • Dashboard: Full Textual TUI with real-time metrics and GPU telemetry (default in interactive terminals)
  • Simple: TQDM progress bars for minimal overhead
  • None: Application logs only, no progress UI (default in non-interactive environments)

All types display the final metrics table upon completion. UI type only affects progress display during benchmark execution.

Note: Both --ui-type and --ui are supported interchangeably.

Automatic TTY Detection

AIPerf automatically detects whether it is running in an interactive terminal (TTY):

  • Interactive terminal (TTY): Defaults to dashboard
  • Non-interactive (piped, redirected, CI/CD): Defaults to none with basic non-rich log formatting

This means aiperf profile ... | tee output.log or aiperf profile ... > output.log will automatically use --ui-type none and plain-text logs without Rich formatting.

To override the automatic detection, explicitly pass --ui-type:

# Force dashboard UI even when piping output
aiperf profile ... --ui-type dashboard | tee output.log

# Force no UI even in an interactive terminal
aiperf profile ... --ui-type none

Dashboard (Default in TTY)

The full-featured TUI provides:

  • Real-time request and record metrics
  • Live GPU telemetry (when enabled)
  • Worker status monitoring
  • Interactive display with multiple tabs
aiperf profile \
  --model Qwen/Qwen3-0.6B \
  --url localhost:8000 \
  --endpoint-type chat \
  --concurrency 10 \
  --request-count 100 \
  --streaming

When to use:

  • Monitoring benchmarks interactively
  • Viewing real-time metrics as they're computed
  • Checking worker status and errors
  • Viewing GPU telemetry

Note: Dashboard automatically switches to simple when using --verbose or --extra-verbose in a TTY for better log visibility.

Simple

Lightweight progress bars using TQDM:

aiperf profile \
  --model Qwen/Qwen3-0.6B \
  --url localhost:8000 \
  --endpoint-type chat \
  --concurrency 10 \
  --request-count 100 \
  --streaming \
  --ui-type simple

Sample Output (Successful Run):

INFO     Starting AIPerf System
INFO     Registered Dataset Manager (id: 'dataset_manager_abc123')
INFO     AIPerf System is PROFILING

Profiling: 100/100 |████████████████████████| 100% [01:45<00:00]
Processing Records: 100/100 |███████████████| 100% [00:01<00:00]

INFO     Benchmark completed successfully
INFO     Results saved to: artifacts/Qwen_Qwen3-0.6B-chat-concurrency10/

When to use:

  • Need lower resource overhead
  • Terminal doesn't support rich TUI rendering
  • Running with verbose logging

None (Default in Non-TTY)

Shows application logs only, no progress UI. This is the automatic default when output is piped or redirected:

aiperf profile \
  --model Qwen/Qwen3-0.6B \
  --url localhost:8000 \
  --endpoint-type chat \
  --concurrency 10 \
  --request-count 100 \
  --streaming \
  --ui-type none

Sample Output (Successful Run):

23:07:28.809795 INFO     Starting AIPerf System
23:07:31.580751 INFO     Registered Dataset Manager (id: 'dataset_manager_abc123')
23:07:31.586297 INFO     Registered Worker Manager (id: 'worker_manager_def456')
23:07:31.595391 INFO     Registered Timing Manager (id: 'timing_manager_ghi789')
23:07:31.987169 INFO     AIPerf System is CONFIGURING
23:07:32.594891 INFO     AIPerf System is CONFIGURED
23:07:32.597896 INFO     AIPerf System is PROFILING
23:09:18.123456 INFO     Benchmark completed successfully
23:09:18.234567 INFO     Results saved to: artifacts/Qwen_Qwen3-0.6B-chat-concurrency10/

When to use:

  • Automating benchmarks in scripts or CI/CD (auto-selected)
  • Piping output to files (auto-selected)
  • Minimizing UI overhead