Multi-GPU with Dask-cuDF#
cuDF is a single-GPU library. For Multi-GPU cuDF solutions we use Dask and the dask-cudf package, which is able to scale cuDF across multiple GPUs on a single machine, or multiple GPUs across many machines in a cluster.
Dask DataFrame was originally designed to scale Pandas, orchestrating many Pandas DataFrames spread across many CPUs into a cohesive parallel DataFrame. Because cuDF currently implements only a subset of the Pandas API, not all Dask DataFrame operations work with cuDF.
The following is tested and expected to work:
What works#
Data ingestion
dask_cudf.read_csvUse standard Dask ingestion with Pandas, then convert to cuDF (For Parquet and other formats this is often decently fast)
Linear operations
Element-wise operations:
df.x + df.y,df ** 2Assignment:
df['z'] = df.x + df.yRow-wise selections:
df[df.x > 0]Loc:
df.loc['2001-01-01': '2005-02-02']Date time/string accessors:
df.timestamp.dt.dayofweek… and most similar operations in this category that are already implemented in cuDF
Reductions
Like
sum,mean,max,count, and so on onSeriesobjectsSupport for reductions on full dataframes
stdCustom reductions with dask.dataframe.reduction
Groupby aggregations
On single columns:
df.groupby('x').y.max()With custom aggregations:
groupby standard deviation
grouping on multiple columns
groupby agg for multiple outputs
Joins:
On full unsorted columns:
left.merge(right, on='id')(expensive)On sorted indexes:
left.merge(right, left_index=True, right_index=True)(fast)On large and small dataframes:
left.merge(cudf_df, on='id')(fast)
Rolling operations
Converting to and from other forms
Dask + Pandas to Dask + cuDF
df.map_partitions(cudf.DataFrame.from_pandas)Dask + cuDF to Dask + Pandas
df.map_partitions(lambda df: df.to_pandas())cuDF to Dask + cuDF:
dask.dataframe.from_pandas(df, npartitions=20)Dask + cuDF to cuDF:
df.compute()
Additionally all generic Dask operations, like compute, persist,
visualize and so on work regardless.
Developing the API#
Above we mention the following:
and most similar operations in this category that are already implemented in cuDF
This is because it is difficult to create a comprehensive list of
operations in the cuDF and Pandas libraries. The API is large enough to
be difficult to track effectively. For any operation that operates
row-wise like fillna or query things will likely, but not
certainly work. If operations don’t work it is often due to a slight
inconsistency between Pandas and cuDF that is generally easy to fix. We
encourage users to look at the cuDF issue
tracker to see if their
issue has already been reported and, if not, raise a new
issue.