agentscope-ai / agentscope-ai/agentscope
环境报错，vllm ，flash_attn 都报错，有没有建议的对应环境版本号
- Dominant language
- Python
- Stars
- 31.5k
- Forks
- 3.5k
- Avg merge
- 1d 23h
- Merged PRs (30d)
- 95
Description
File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/transformers/modeling_utils.py", line 2812, in _check_and_adjust_attn_implementation
[36m(Trainer pid=2756877)[0m lazy_import_flash_attention(applicable_attn_implementation)
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/transformers/modeling_flash_attention_utils.py", line 136, in lazy_import_flash_attention
[36m(Trainer pid=2756877)[0m _flash_fn, _flash_varlen_fn, _pad_fn, _unpad_fn = _lazy_imports(implementation)
[36m(Trainer pid=2756877)[0m ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/transformers/modeling_flash_attention_utils.py", line 83, in _lazy_imports
[36m(Trainer pid=2756877)[0m from flash_attn import flash_attn_func, flash_attn_varlen_func
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/flash_attn/__init__.py", line 3, in
[36m(Trainer pid=2756877)[0m from flash_attn.flash_attn_interface import (
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/flash_attn/flash_attn_interface.py", line 15, in
[36m(Trainer pid=2756877)[0m import flash_attn_2_cuda as flash_attn_gpu
[36m(Trainer pid=2756877)[0m ImportError: /home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/flash_attn_2_cuda.cpython-312-x86_64-linux-gnu.so: undefined symbol: _ZN3c104cuda29c10_cuda_check_implementationEiPKcS2_ib
[36m(Trainer pid=2756877)[0m Unhandled error (suppress with 'RAY_IGNORE_UNHANDLED_ERRORS=1'): [36mray::WorkerDict.ref_init_model()[39m (pid=2757296, ip=10.21.175.69, actor_id=eba58008e2387b9be7d855ce12000000, repr=)
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/verl/single_controller/ray/base.py", line 841, in func
[36m(Trainer pid=2756877)[0m return getattr(self.worker_dict[key], name)(*args, **kwargs)
[36m(Trainer pid=2756877)[0m ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/verl/single_controller/base/decorator.py", line 456, in inner
[36m(Trainer pid=2756877)[0m return func(*args, **kwargs)
File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/concurrent/futures/_base.py", line 401, in __get_result[32m [repeated 4x across cluster][0m
[36m(Trainer pid=2756877)[0m raise self._exception[32m [repeated 4x across cluster][0m
[36m(Trainer pid=2756877)[0m ^^^^^^^^^^^^^^^^^^^^^[32m [repeated 21x across cluster][0m
[36m(Trainer pid=2756877)[0m ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^[32m [repeated 8x across cluster][0m
[36m(Trainer pid=2756877)[0m ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^[32m [repeated 8x across cluster][0m
[36m(Trainer pid=2756877)[0m File "/home/zhangkailin/.conda/envs/agent/lib/python3.12/site-packages/transformers/modeling_utils.py", line 2197, in __init__[32m [repeated 13x across cluster][0m
[36m(vLLMRolloutModel pid=2757291)[0m self.default_sampling_params = vllm.SamplingParams([32m [repeated 3x across cluster][0m
[36m(vLLMRolloutModel pid=2757291)[0m ^^^^^^^^^^^^^^^^^^^^[32m [repeated 3x across cluster][0m
[36m(vLLMRolloutModel pid=2757291)[0m TypeError: Unexpected keyword argument 'truncate_prompt_tokens'[32m [repeated 3x across cluster][0m
Contributor guide
Assessment
This issue has not been assessed yet.