Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

latency test enhancement - part 1 #909

Merged
merged 4 commits into from
Aug 4, 2024
Merged

latency test enhancement - part 1 #909

merged 4 commits into from
Aug 4, 2024

Conversation

min-xu-et
Copy link
Contributor

@min-xu-et min-xu-et commented Aug 4, 2024 •
edited
Loading

This is part 1 of change to make latency test generate graph.

  • this makes the batch_size arg a list, but user can continue use it the same way as before when specifying a single value
  • this optionally dumps the results in a json file. example:
$ cat out.json
[
  {
    "batch_size": 1,
    "output_len": 256,
    "prefill_latency": 0.12703561782836914,
    "prefill_throughput": 4030.3657253962833,
    "avg_decode_latency": 0.05401019565761089,
    "avg_decode_throughput": 18.515022725326567,
    "total_latency": 13.953645706176758,
    "total_throughput": 55.03937939745991
  },
  {
    "batch_size": 1,
    "output_len": 256,
    "prefill_latency": 0.137282133102417,
    "prefill_throughput": 3729.545778677777,
    "avg_decode_latency": 0.054004695266485214,
    "avg_decode_throughput": 18.516908484818174,
    "total_latency": 13.962484121322632,
    "total_throughput": 55.00453882895798
  }
]

merrymercy
merrymercy requested changes Aug 4, 2024
View reviewed changes
python/sglang/bench_latency.py Show resolved Hide resolved
python/sglang/bench_latency.py Outdated Show resolved Hide resolved
@min-xu-et
use jsonl format
f52110d 
- testing output

```
In [3]: with open("out.jsonl", "r") as f:
   ...:     for l in f.readlines():
   ...:         print(json.loads(l))
   ...:         print("-"*80)
   ...:
{'batch_size': 1, 'output_len': 256, 'prefill_latency': 0.13420748710632324, 'prefill_throughput': 3814.9883515394195, 'avg_decode_latency': 0.053959593176841736, 'avg_decode_throughput': 18.532385830313817, 'total_latency': 13.947863340377808, 'total_throughput': 55.06219707334737}
--------------------------------------------------------------------------------
{'batch_size': 1, 'output_len': 256, 'prefill_latency': 0.13029885292053223, 'prefill_throughput': 3929.4282990523557, 'avg_decode_latency': 0.05398500803858042, 'avg_decode_throughput': 18.523661222488833, 'total_latency': 13.95046091079712, 'total_throughput': 55.051944513575}
--------------------------------------------------------------------------------
{'batch_size': 1, 'output_len': 256, 'prefill_latency': 0.13561320304870605, 'prefill_throughput': 3775.4436034971686, 'avg_decode_latency': 0.05398557987064123, 'avg_decode_throughput': 18.52346501410511, 'total_latency': 13.955921649932861, 'total_throughput': 55.03040352793143}
--------------------------------------------------------------------------------
```
@min-xu-et min-xu-et requested a review from merrymercy August 4, 2024 04:54
@min-xu-et
use jsonlines library
bcfafd3 
- simpler code
@Ying1123 Ying1123 merged commit 5398564 into sgl-project:main Aug 4, 2024
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Ying1123 Ying1123 Ying1123 approved these changes

@merrymercy merrymercy Awaiting requested review from merrymercy

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@min-xu-et @Ying1123 @merrymercy