change llm benchmark seqlen from 1024 to 256

[nihui]

No description provided.

[tencent-adm]

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[Copilot]

Pull request overview

Updates the LLM benchmarking harness to use a shorter prefill/past sequence length (256 instead of 1024), likely to reduce runtime and resource usage during benchmark runs.

Changes:

• Switch decoder warmup/prefill benchmark sequence length from 1024 to 256.
• Switch decode benchmark past sequence length from 1024 to 256.
• Update benchmark case name suffixes from 1k to 256 for clearer output labeling.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.