GPU Support

Polars provides an in-memory, GPU-accelerated execution engine for the Lazy API in Python using RAPIDS cuDF on NVIDIA GPUs. This functionality is available in Open Beta, is undergoing rapid development, and is currently a single GPU implementation.

If you install Polars with the GPU feature flag, you can trigger GPU-based execution by running .collect(engine="gpu") instead of .collect().

Python

import polars as pl

df = pl.LazyFrame({"a": [1.242, 1.535]})

q = df.select(pl.col("a").round(1))

result = q.collect(engine="gpu")
print(result)

shape: (2, 1)
┌─────┐
│ a   │
│ --- │
│ f64 │
╞═════╡
│ 1.2 │
│ 1.5 │
└─────┘

Learn more in the GPU Support guide.