Simple Xarray + JAX Integration

This is an experiment at integrating Xarray + JAX in a simple way, leveraging equinox.

import jax.numpy as jnp
import xarray as xr
import xarray_jax as xj

# Construct a DataArray.
da = xr.DataArray(
    xr.Variable(["x", "y"], jnp.ones((2, 3))),
    coords={"x": [1, 2], "y": [3, 4, 5]},
    name="foo",
    attrs={"attr1": "value1"},
)

# Do some operations inside a JIT compiled function.
@eqx.filter_jit
def some_function(data):
    neg_data = -1.0 * data
    return neg_data * neg_data.coords["y"] # Multiply data by coords.

da = some_function(da)

# Construct a xr.DataArray with dummy data (useful for tree manipulation).
da_mask = jax.tree.map(lambda _: True, data)

# Use jax.grad.
@eqx.filter_jit
def fn(data):
    return (data**2.0).sum().data

grad = jax.grad(fn)(da)

# Convert to a custom XjDataArray, implemented as an equinox module.
# (Useful for avoiding potentially weird xarray interactions with JAX).
xj_da = xj.from_xarray(da)

# Convert back to a xr.DataArray.
da = xj.to_xarray(xj_da)

Installation

pip install xarray_jax

Status

• PyTree node registrations
  • xr.Variable
  • xr.DataArray
  • xr.Dataset
• Minimal shadow types implemented as equinox modules to handle edge cases (Note: these types are merely data structures that contain the data of these types. They don't have any of the methods of the xarray types).
  • XjVariable
  • XjDataArray
  • XjDataset
• xj.from_xarray and xj.to_xarray functions to go between xj and xr types.
• Support for xr types with dummy data (useful for tree manipulation).
• Support for transformations that change the dimensionality of the data.

Sharp Edges

Prefer eqx.filter_jit over jax.jit

There are some edge cases with metadata that eqx.filter_jit handles but jax.jit does not.

Operations that Increase the Dimensionality of the Data

Operations that increase the dimensionality of the data (e.g. jnp.expand_dims) will cause problems downstream.

var = xr.Variable(dims=("x", "y"), data=jnp.ones((3, 3)))

# This will not error.
var = jax.tree.map(lambda x: jnp.expand_dims(x, axis=0), var)

# The error from expanding the dimensionality will be triggered here.
var = var + 1 

Dispatching to jnp is not supported yet

Pending resolution of https://github.com/pydata/xarray/issues/7848.

var = xr.Variable(dims=("x", "y"), data=jnp.ones((3, 3)))

# This will fail.
jnp.square(var)

# This will work.
xr.apply_ufunc(jnp.square, var)

Distinction from the GraphCast Implementation

This experiment is largely inspired by the GraphCast implementation, with a direct re-use of the _HashableCoords in that project.

However, this experiment aims to:

1. Take a more minimialist approach (and thus neglects some features such as support JAX arrays as coordinates).
2. Find a solution more compatible with common JAX PyTree manipulation patterns that trigger errors with Xarray types. For example, it's common to use boolean masks to filter out elements of a PyTree, but this tends to fail with Xarray types.

Acknowledgements

This repo was made possible by great discussions within the JAX + Xarray open source community, especially this one. In particular, the author would like to acknowledge @shoyer, @mjwillson, and @TomNicholas.