[Bug] Lora Non-Flashinfer backend select error

[BruceXcluding]

Checklist

• 1. I have searched related issues but cannot get the expected help.
• 2. The bug has not been fixed in the latest version.
• 3. Please note that if the bug-related issue you submitted lacks corresponding environment info and a minimal reproducible demo, it will be challenging for us to reproduce and resolve the issue, reducing the likelihood of receiving feedback.
• 4. If the issue you raised is not a bug but a question, please raise a discussion at https://github.com/sgl-project/sglang/discussions/new/choose Otherwise, it will be closed.
• 5. Please use English, otherwise it will be closed.

Describe the bug

Lora Backend Error

Bug report:

INFO 02-05 07:04:11 __init__.py:179] Automatically detected platform rocm.
WARNING 02-05 07:04:11 rocm.py:34] `fork` method is not supported by ROCm. VLLM_WORKER_MULTIPROC_METHOD is overridden to `spawn` instead.
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/sgl-workspace/sglang_public/python/sglang/bench_one_batch.py", line 60, in <module>
    from sglang.srt.entrypoints.engine import _set_envs_and_config
  File "/sgl-workspace/sglang_public/python/sglang/srt/entrypoints/engine.py", line 36, in <module>
    from sglang.srt.managers.data_parallel_controller import (
  File "/sgl-workspace/sglang_public/python/sglang/srt/managers/data_parallel_controller.py", line 31, in <module>
    from sglang.srt.managers.scheduler import run_scheduler_process
  File "/sgl-workspace/sglang_public/python/sglang/srt/managers/scheduler.py", line 81, in <module>
    from sglang.srt.managers.tp_worker import TpModelWorker
  File "/sgl-workspace/sglang_public/python/sglang/srt/managers/tp_worker.py", line 31, in <module>
    from sglang.srt.model_executor.model_runner import ModelRunner
  File "/sgl-workspace/sglang_public/python/sglang/srt/model_executor/model_runner.py", line 47, in <module>
    from sglang.srt.lora.lora_manager import LoRAManager
  File "/sgl-workspace/sglang_public/python/sglang/srt/lora/lora_manager.py", line 23, in <module>
    from sglang.srt.lora.backend import TritonLoraBackend
  File "/sgl-workspace/sglang_public/python/sglang/srt/lora/backend/__init__.py", line 2, in <module>
    from .flashinfer_backend import FlashInferLoraBackend
  File "/sgl-workspace/sglang_public/python/sglang/srt/lora/backend/flashinfer_backend.py", line 4, in <module>
    from flashinfer import SegmentGEMMWrapper
ModuleNotFoundError: No module named 'flashinfer'

Issue PR: [Feature] Define backends and add Triton backend for Lora (#3161)
Issue Details:
https://github.com/sgl-project/sglang/blob/main/python/sglang/srt/lora/backend/__init__.py
https://github.com/sgl-project/sglang/blob/main/python/sglang/srt/lora/lora_manager.py#L23
This is not a selection; it is with flashinfer as well.
work with:

__init__:
#from .flashinfer_backend import FlashInferLoraBackend

__all__ = [
    #"FlashInferLoraBackend",
    "TritonLoraBackend",
]

lora_manager:
from sglang.srt.lora.backend import FlashInferLoraBackend, TritonLoraBackend -> from sglang.srt.lora.backend import TritonLoraBackend

Reproduction

python -m sglang.bench_one_batch --batch-size 32 --input 128 --output 32 --model deepseek-ai/DeepSeek-V3 --tp 8 --trust-remote-code

Environment

ROCM with latest code:

git clone https://github.com/sgl-project/sglang.git
cd sglang

pip install --upgrade pip
cd sgl-kernel
python setup_rocm.py install
cd ..
pip install -e "python[all_hip]"

[Fridge003]

@BruceXcluding Hi, this bug should have been fixed in #3312. Would you please try again.

[BruceXcluding]

@BruceXcluding Hi, this bug should have been fixed in #3312. Would you please try again.

Fixed. Thank you