Skip to content

Conversation

@herotai214
Copy link
Collaborator

@herotai214 herotai214 commented Dec 5, 2025

[updates 1217]

============ Serving Benchmark Result ============
Successful requests: 100
Failed requests: 0
Benchmark duration (s): 13.40
Total input tokens: 25524
Total generated tokens: 10656
Request throughput (req/s): 7.46
Output token throughput (tok/s): 795.07
Peak output token throughput (tok/s): 2420.00
Peak concurrent requests: 100.00
Total Token throughput (tok/s): 2699.49
---------------Time to First Token----------------
Mean TTFT (ms): 6366.42
Median TTFT (ms): 6819.96
P99 TTFT (ms): 10187.31
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms): 45.11
Median TPOT (ms): 42.54
P99 TPOT (ms): 110.58
---------------Inter-token Latency----------------
Mean ITL (ms): 53.41
Median ITL (ms): 25.41
P99 ITL (ms): 602.57
==================================================


WORKS; accuracy ~0.8.
haven't edit or test code to work properly without EPD or without nixl.

Purpose

Test Plan

Test Result


Essential Elements of an Effective PR Description Checklist
  • The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
  • The test plan, such as providing test command.
  • The test results, such as pasting the results comparison before and after, or e2e results
  • (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
  • (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant