You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
import dask
from dask.delayed import delayed
from dask.distributed import Client
from dask_cuda import LocalCUDACluster
cluster = LocalCUDACluster()
Exception:
The interesting part:
TypeError: __init__() got an unexpected keyword argument 'env'
/opt/anaconda3/lib/python3.7/site-packages/dask_cuda-0.0.0-py3.7.egg/dask_cuda/local_cuda_cluster.py in _start(self, ip, n_workers)
77 env={"CUDA_VISIBLE_DEVICES": cuda_visible_devices(i)},
78 )
---> 79 for i in range(n_workers)
80 ]
81
More of the code around the exception for context:
@gen.coroutine
def _start(self, ip=None, n_workers=0):
"""
Start all cluster services.
"""
if self.status == "running":
return
if (ip is None) and (not self.scheduler_port) and (not self.processes):
# Use inproc transport for optimization
scheduler_address = "inproc://"
elif ip is not None and ip.startswith("tls://"):
scheduler_address = "%s:%d" % (ip, self.scheduler_port)
else:
if ip is None:
ip = "127.0.0.1"
scheduler_address = (ip, self.scheduler_port)
self.scheduler.start(scheduler_address)
yield [
self._start_worker(
**self.worker_kwargs,
env={"CUDA_VISIBLE_DEVICES": cuda_visible_devices(i)},
)
for i in range(n_workers)
]
The function signature for _start_worker doesn't have an env argument, though I'm not sure why it isn't absorbed by **kwargs. (also why call _start_work when there is a public function start_worker?).
Software:
Anaconda3 2018.12 (Python 3.7)
Rapids 0.5.0 - pip install of cudf-cuda100
dask-cuda - git commit e405194.
notebook: git commit dcbc0d9238e86a0a5383f6841b3f13223bd0bebb notebooks/dask/Dask_Hello_World.ipynb
Exception:
The interesting part:
More of the code around the exception for context:
The function signature for _start_worker doesn't have an env argument, though I'm not sure why it isn't absorbed by **kwargs. (also why call _start_work when there is a public function start_worker?).
The full exception:
The text was updated successfully, but these errors were encountered: