[HiSparse] Clarify decode token usage logs

[huangtingwei9988]

Motivation

before

[2026-04-08 14:05:21 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 144.95, #queue-req: 0
[2026-04-08 14:05:22 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 140.27, #queue-req: 0
[2026-04-08 14:05:24 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 136.39, #queue-req: 0
[2026-04-08 14:05:25 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 139.89, #queue-req: 0
[2026-04-08 14:05:27 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 144.10, #queue-req: 0
[2026-04-08 14:05:28 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 157.64, #queue-req: 0
[2026-04-08 14:05:29 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 153.78, #queue-req: 0
[2026-04-08 14:05:30 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 157.11, #queue-req: 0
[2026-04-08 14:05:32 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 154.96, #queue-req: 0
[2026-04-08 14:05:33 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 152.94, #queue-req: 0
[2026-04-08 14:05:34 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 154.59, #queue-req: 0
[2026-04-08 14:05:36 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 159.42, #queue-req: 0
[2026-04-08 14:05:37 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 159.05, #queue-req: 0
[2026-04-08 14:05:38 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 154.52, #queue-req: 0
[2026-04-08 14:05:39 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 157.74, #queue-req: 0
[2026-04-08 14:05:41 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 159.45, #queue-req: 0
[2026-04-08 14:05:42 TP0] Decode batch, #running-req: 5, #token: 20800, token usage: 0.29, cuda graph: True, gen throughput (token/s): 162.58, #queue-req: 0

after

[2026-04-08 14:09:56 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 41422, cpu token usage: 0.29, cuda graph: True, gen throughput (token/s): 153.75, #queue-req: 0
[2026-04-08 14:09:58 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 41622, cpu token usage: 0.29, cuda graph: True, gen throughput (token/s): 155.19, #queue-req: 0
[2026-04-08 14:09:59 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 41822, cpu token usage: 0.29, cuda graph: True, gen throughput (token/s): 151.59, #queue-req: 0
[2026-04-08 14:10:00 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 42022, cpu token usage: 0.29, cuda graph: True, gen throughput (token/s): 154.91, #queue-req: 0
[2026-04-08 14:10:02 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 42222, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 153.62, #queue-req: 0
[2026-04-08 14:10:03 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 42422, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 151.58, #queue-req: 0
[2026-04-08 14:10:04 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 42622, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 153.37, #queue-req: 0
[2026-04-08 14:10:06 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 42822, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 139.67, #queue-req: 0
[2026-04-08 14:10:07 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 43022, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 138.92, #queue-req: 0
[2026-04-08 14:10:08 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 43222, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 140.55, #queue-req: 0
[2026-04-08 14:10:10 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 43422, cpu token usage: 0.30, cuda graph: True, gen throughput (token/s): 155.25, #queue-req: 0
[2026-04-08 14:10:11 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 43622, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 156.24, #queue-req: 0
[2026-04-08 14:10:12 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 43822, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 155.15, #queue-req: 0
[2026-04-08 14:10:14 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 44022, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 154.88, #queue-req: 0
[2026-04-08 14:10:15 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 44222, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 155.86, #queue-req: 0
[2026-04-08 14:10:16 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 44422, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 156.36, #queue-req: 0
[2026-04-08 14:10:17 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 44622, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 157.12, #queue-req: 0
[2026-04-08 14:10:19 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 44822, cpu token usage: 0.31, cuda graph: True, gen throughput (token/s): 157.10, #queue-req: 0
[2026-04-08 14:10:20 TP0] Decode batch, #running-req: 5, #gpu-token: 20800, gpu token usage: 0.29, #cpu-token: 45022, cpu token usage: 0.32, cuda graph: True, gen throughput (token/s): 146.91, #queue-req: 0

Modifications

Accuracy Tests

Speed Tests and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
• Follow the SGLang code style guidance.

Review and Merge Process

1. Ping Merge Oncalls to start the process. See the PR Merge Process.
2. Get approvals from CODEOWNERS and other reviewers.
3. Trigger CI tests with comments or contact authorized users to do so.
   • Common commands include /tag-and-rerun-ci, /tag-run-ci-label, /rerun-failed-ci
4. After green CI and required approvals, ask Merge Oncalls or people with Write permission to merge the PR.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[huangtingwei9988]

/tag-and-rerun-ci

[huangtingwei9988]

/rerun-failed-ci

[huangtingwei9988]

/rerun-failed-ci

[hzh0425]

@huangtingwei9988 pls resolve the conflict

[hzh0425]

/rerun-test test_dsa_models_hisparse.py

[github-actions]

❌ test_dsa_models_hisparse.py: No test file found matching test_dsa_models_hisparse.py under test/registered/.

[hzh0425]

/rerun-test test_dsa_models_hisparse.py

[github-actions]

✅ 8-gpu-h200 (1 test): View workflow run

cd test/ && python3 registered/8-gpu-models/test_dsa_models_hisparse.py