Turn off PREBUILD aiter in MI355

[1am9trash]

Motivation

There are some errors in aiter PREBUILD mode, which cause accuracy drop to 0% on DS-MXFP4 model. The errors have been solved but not yet merge into aiter main. We may use jit build to make the docker run correctly for now.

• Prebuild kernel dispatching issue: fixed in fix_fp4_aot_post2.
• MLA issue: fixed in mla_fix.

Modifications

Turn off BUILD_AITER_ALL flag in rocm.dockerfile.

Accuracy Tests

Build the image locally, and then run sglang server with jit build.
Machine: MI355 * 8
Model: DeepSeek-R1-MXFP4-Preview

root@smci355-ccs-aus-m15-21:/sgl-workspace/sglang# python3 benchmark/gsm8k/bench_sglang.py --num-questions 2000 --parallel 2000 --port 8000
100%|██████████████████████████████████████████████████████████████| 1319/1319 [01:58<00:00, 11.14it/s]
Accuracy: 0.937
Invalid: 0.000
Latency: 118.575 s

Benchmarking 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.
• Work with maintainers to merge your PR. See the PR Merge Process

[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!