[dask-on-ray] ValueError on read-only memory #10124
Description
What is the problem?
Ray version and other system information (Python version, TensorFlow version, OS): 0.9dev
I was trying out the new Ray backend plugin for Dask and ran into a read-only error. Looks like the Dask code tries to write in-place, but Ray doesn't allow it since objects are immutable.
Reproduction (REQUIRED)
Please provide a script that can be run to reproduce the issue. The script should have no external library dependencies (i.e., use fake or mock data / environments):
import ray
import dask
import dask.dataframe as dd
import pandas as pd
import numpy as np
from ray.experimental.dask import ray_dask_get
import time
N_ROWS = 1000
N_PARTITIONS = 100
ray.init()
df = dd.from_pandas(pd.DataFrame(np.random.randn(N_ROWS, 2), columns=['a', 'b']), npartitions=N_PARTITIONS)
start = time.time()
print(df.groupby('b').a.sum().compute(scheduler=ray_dask_get))
end = time.time()
print("ray", end - start)
If we cannot run your script, we cannot fix your issue.
- I have verified my script runs in a clean environment and reproduces the issue.
- I have verified the issue also occurs with the latest wheels.