Run tests using JAX as the backend array API (on CPU) (#374)

.github/workflows/jax-tests.yml

@@ -0,0 +1,49 @@
+name: JAX tests
+
+on:
+  schedule:
+    # Every weekday at 03:53 UTC, see https://crontab.guru/
+    - cron: "53 3 * * 1-5"
+  workflow_dispatch:
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  test:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: ["ubuntu-latest"]
+        python-version: ["3.9"]
+
+    steps:
+      - name: Checkout source
+        uses: actions/checkout@v3
+        with:
+          fetch-depth: 0
+
+      - name: Set up Python
+        uses: actions/setup-python@v3
+        with:
+          python-version: ${{ matrix.python-version }}
+          architecture: x64
+
+      - name: Setup Graphviz
+        uses: ts-graphviz/setup-graphviz@v1
+
+      - name: Install
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install -e '.[test]' 'jax[cpu]'
+          python -m pip uninstall -y lithops  # tests don't run on Lithops
+
+      - name: Run tests
+        run: |
+          # exclude tests that rely on structured types since JAX doesn't support these
+          pytest -k "not argmax and not argmin and not mean and not apply_reduction and not broadcast_trick"
+        env:
+          CUBED_BACKEND_ARRAY_API_MODULE: jax.numpy
+          JAX_ENABLE_X64: True

cubed/backend_array_api.py

@@ -1,13 +1,38 @@
+import os
+from importlib import import_module
+
 import numpy as np
 
-# The array implementation used for backend operations.
-# This must be compatible with the Python Array API standard, although
+# The array implementation used for backend operations is stored in the
+# namespace variable, and defaults to array_api_compat.nump, unless it
+# is overridden by an environment variable.
+# It must be compatible with the Python Array API standard, although
 # some extra functions are used too (nan functions, take_along_axis),
 # which array_api_compat provides, but other Array API implementations
 # may not.
-import array_api_compat.numpy  # noqa: F401 isort:skip
 
-namespace = array_api_compat.numpy
+if "CUBED_BACKEND_ARRAY_API_MODULE" in os.environ:
+    # This code is based on similar code in array_api_tests
+    xp_name = os.environ["CUBED_BACKEND_ARRAY_API_MODULE"]
+    _module, _sub = xp_name, None
+    if "." in xp_name:
+        _module, _sub = xp_name.split(".", 1)
+    xp = import_module(_module)
+    if _sub:
+        try:
+            xp = getattr(xp, _sub)
+        except AttributeError:
+            # _sub may be a submodule that needs to be imported. WE can't
+            # do this in every case because some array modules are not
+            # submodules that can be imported (like mxnet.nd).
+            xp = import_module(xp_name)
+    namespace = xp
+
+else:
+    import array_api_compat.numpy
+
+    namespace = array_api_compat.numpy
+
 
 # These functions to convert to/from backend arrays
 # assume that no extra memory is allocated, by using the

cubed/storage/virtual.py

@@ -5,6 +5,7 @@
 import zarr
 from zarr.indexing import BasicIndexer, is_slice
 
+from cubed.backend_array_api import backend_array_to_numpy_array
 from cubed.backend_array_api import namespace as nxp
 from cubed.backend_array_api import numpy_array_to_backend_array
 from cubed.types import T_DType, T_RegularChunks, T_Shape
@@ -107,7 +108,7 @@ class VirtualInMemoryArray:
 
     def __init__(
         self,
-        array: np.ndarray,  # TODO: generalise
+        array: np.ndarray,  # TODO: generalise to array API type
         chunks: T_RegularChunks,
         max_nbytes: int = 10**6,
     ):
@@ -129,7 +130,7 @@ def __init__(
         self.chunks = template.chunks
         self.template = template
         if array.size > 0:
-            template[...] = array
+            template[...] = backend_array_to_numpy_array(array)
 
     def __getitem__(self, key):
         return self.array.__getitem__(key)