MyYinYin

Could you tell me the environment of your expriment, such as the version of tensorflow, python or cuda

installation is too long

2

cuda:12.1,torch:2.2.1 , vllm:0.6.7 , transformers : 4.37.2, when I run：pip install flash-attn==2.5.6 --no-build-isolation --extra-index-url https://pypi.nvidia.com --no-deps， it looks like being stuck. [root@172-24-168-106 quick_start]$pip install flash-attn==2.5.6 --no-build-isolation --extra-index-url https://pypi.nvidia.com --no-deps Looking...

ModuleNotFoundError: No module named 'transformers.modeling_rope_utils'

streamlit run demo_en.py --server.port 6006 --server.address 0.0.0.0报错

1

我在运行streamlit run demo_en.py --server.port 6006 --server.address 0.0.0.0这个代码时，报这个错：ImportError: cannot import name '_get_default_timeout' from 'torch.distributed.distributed_c10d' (/opt/conda/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py),我的troch是2.1.0，cuda是11.8，vllm是0.8.0，flash-attn是2.2.0