Halogen monitor
connecting…
Throughput
–tok/s live
Active requests
–
generating right now
Waiting in queue
–
Last 5 minutes
–tokens
CPU
–
RAM
–
GPU load
–
GPU memory (GTT)
–
Throughput, last 15 min (tok/s)
live tok/s
Active vs waiting, last 15 min
activewaiting
Recent completed requests
| finished | tokens | duration | tok/s | prompt | cached |
|---|
Services
Table view of the charts (last 20 samples)
| time | tok/s | active | waiting | cpu % | gpu % |
|---|
Numbers are estimated from the engine/API logs: live tok/s sums the active requests' latest rate; queue depth comes from the API's "waited for a slot" messages.