[Hyperloom] Optimize dsr1-fp8-mi355x-sglang, gptoss-fp4-mi355x-vllm

[lishuoshuo-amd]

Description

Automated performance optimization update from Hyperloom CI.

dsr1-fp8-mi355x-sglang

Metric	Value
Baseline (tok/s/GPU)	309.88
Optimized (tok/s/GPU)	331.37
Optimization Gain	+6.9%
InferenceX Current (tok/s/GPU)	310.09
vs InferenceX	+6.9%

Server flag changes:

• --num-continuous-decode-steps: 4 → 8

gptoss-fp4-mi355x-vllm

Metric	Value
Baseline (tok/s/GPU)	7344.93
Optimized (tok/s/GPU)	7855.14
Optimization Gain	+7.0%
InferenceX Current (tok/s/GPU)	6585.93
vs InferenceX	+19.3%

Server flag changes:

• Add --max-num-seqs 256
• Add --enable-chunked-prefill
• Add --max-num-batched-tokens 16384

Related Issue

Automated by Hyperloom CI

Type of Change

• Configuration change

Checklist

• I have tested my changes locally
• I have updated documentation if necessary
• If I changed a container image or config, I have already updated perf-changelog.yaml