提问后出现错误提示 AttributeError: module 'eventlet.green.select' has no attribute 'epoll'

[leeeex]

kagentsys --query="刘德华老婆是谁？" --llm_name="kagentlms_qwen_7b_mat" --use_local_llm --local_llm_host="localhost" --local_llm_port=8899 --lang="zh"

File "/opt/conda/lib/python3.10/site-packages/trio-0.24.0-py3.10.egg/trio/_core/_io_epoll.py", line 203, in EpollIOManager
_epoll: select.epoll = attr.ib(factory=select.epoll)
AttributeError: module 'eventlet.green.select' has no attribute 'epoll'

---

还有就是：单卡模型推理服务 ，加载模型后隔一小段时间后就会崩溃，然后出现以下提示
ValueError: The model's max seq len (8192) is larger than the maximum number of tokens that can be stored in KV cache (5392). Try increasing gpu_memory_utilization or decreasing max_model_len when initializing the engine.

运行的环境是魔搭阿里云DSW，T_T。。。求大佬帮忙解决

[leeeex]

模型是从魔搭自动下载的，单卡模型推理服务 启动后的日志
root@dsw-279718-6868cc44f8-52b8b:/mnt/workspace# python -m fastchat.serve.vllm_worker --model-path KwaiKEG/kagentlms_qwen_7b_mat --trust-remote-code --dtype half
2024-02-04 17:37:34 | ERROR | stderr | 2024-02-04 17:37:34,809 - modelscope - INFO - PyTorch version 2.1.2+cu121 Found.
2024-02-04 17:37:34 | ERROR | stderr | 2024-02-04 17:37:34,811 - modelscope - INFO - TensorFlow version 2.14.0 Found.
2024-02-04 17:37:34 | ERROR | stderr | 2024-02-04 17:37:34,811 - modelscope - INFO - Loading ast index from /mnt/workspace/.cache/modelscope/ast_indexer
2024-02-04 17:37:34 | ERROR | stderr | 2024-02-04 17:37:34,992 - modelscope - INFO - Loading done! Current index file version is 1.11.0, with md5 23d6ca2be1734e5d30a45b75a5442622 and a total number of 953 components indexed
WARNING 02-04 17:37:35 config.py:457] Casting torch.bfloat16 to torch.float16.
INFO 02-04 17:37:35 llm_engine.py:70] Initializing an LLM engine with config: model='/mnt/workspace/.cache/modelscope/KwaiKEG/kagentlms_qwen_7b_mat', tokenizer='/mnt/workspace/.cache/modelscope/KwaiKEG/kagentlms_qwen_7b_mat', tokenizer_mode=auto, revision=None, tokenizer_revision=None, trust_remote_code=True, dtype=torch.float16, max_seq_len=8192, download_dir='/mnt/workspace/.cache/modelscope/KwaiKEG/kagentlms_qwen_7b_mat', load_format=auto, tensor_parallel_size=1, quantization=None, enforce_eager=False, seed=0)
WARNING 02-04 17:37:36 tokenizer.py:62] Using a slow tokenizer. This might cause a significant slowdown. Consider using a fast tokenizer instead.