使用mindformer跑官方仓储的llama 7b , modelarts的notebook 没有问题正常训练,提交训练作业出现报错(都为8卡,完全一样的镜像 mindformer v0.8):
Traceback (most recent call last):
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/run_mindformer.py", line 340, in <module>
main(config_)
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/tools/cloud_adapter/cloud_monitor.py", line 43, in wrapper
raise exc
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/tools/cloud_adapter/cloud_monitor.py", line 33, in wrapper
result = run_func(*args, **kwargs)
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/run_mindformer.py", line 84, in main
build_context(config)
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/core/context/build_context.py", line 42, in build_context
local_rank, device_num = init_context(use_parallel=config.use_parallel,
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/core/context/build_context.py", line 93, in init_context
init()
File "/home/ma-user/anaconda3/envs/MindSpore/lib/python3.9/site-packages/mindspore/communication/management.py", line 171, in init
init_hccl()
RuntimeError: Ascend collective communication initialization failed.
----------------------------------------------------
- Framework Error Message: (For framework developers)
----------------------------------------------------
Init hccl graph adapter failed.
----------------------------------------------------
- C++ Call Stack: (For framework developers)
----------------------------------------------------
mindspore/ccsrc/plugin/device/ascend/hal/hardware/ascend_collective_comm_lib.cc:129 Initialize
mindspore/ccsrc/plugin/device/ascend/hal/hccl_adapter/hccl_adapter.cc:452 InitKernelInfoStore
Traceback (most recent call last):
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/run_mindformer.py", line 340, in <module>
main(config_)
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/tools/cloud_adapter/cloud_monitor.py", line 43, in wrapper
raise exc
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/tools/cloud_adapter/cloud_monitor.py", line 33, in wrapper
result = run_func(*args, **kwargs)
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/run_mindformer.py", line 84, in main
build_context(config)
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/core/context/build_context.py", line 42, in build_context
local_rank, device_num = init_context(use_parallel=config.use_parallel,
File "/home/ma-user/modelarts/user-job-dir/mindformers/scripts/mf_parallel7/mindformers/core/context/build_context.py", line 93, in init_context
init()
File "/home/ma-user/anaconda3/envs/MindSpore/lib/python3.9/site-packages/mindspore/communication/management.py", line 171, in init
init_hccl()
RuntimeError: Ascend collective communication initialization failed.
----------------------------------------------------
- Framework Error Message: (For framework developers)
----------------------------------------------------
Init hccl graph adapter failed.
----------------------------------------------------
- C++ Call Stack: (For framework developers)
----------------------------------------------------
mindspore/ccsrc/plugin/device/ascend/hal/hardware/ascend_collective_comm_lib.cc:129 Initialize
mindspore/ccsrc/plugin/device/ascend/hal/hccl_adapter/hccl_adapter.cc:452 InitKernelInfoStore