Convert Pandas dataframe to Dask dataframe

I think you can use dask.dataframe.from_pandas:

from dask import dataframe as dd 
sd = dd.from_pandas(df, npartitions=3)
print (sd)
dd.DataFrame<from_pa..., npartitions=2, divisions=(0, 1, 2)>

EDIT:

I find solution:

import pandas as pd
import dask.dataframe as dd
from dask.dataframe.utils import make_meta

df=pd.DataFrame({'a':[1,2,3],'b':[4,5,6]})

dsk = {('x', 0): df}

meta = make_meta({'a': 'i8', 'b': 'i8'}, index=pd.Index([], 'i8'))
d = dd.DataFrame(dsk, name="x", meta=meta, divisions=[0, 1, 2])
print (d)
dd.DataFrame<x, npartitions=2, divisions=(0, 1, 2)>