Update to jax 0.4.19

.github/action/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvidia/cuda:11.8.0-devel-ubuntu20.04
+FROM nvidia/cuda:11.8.0-devel-ubuntu22.04
 
 RUN apt-get update && \
     DEBIAN_FRONTEND=noninteractive apt-get install -y git python3-pip cmake

.github/workflows/tests.yml

@@ -21,9 +21,9 @@ jobs:
           fetch-depth: 0
 
       - name: Set up Python
-        uses: actions/setup-python@v1
+        uses: actions/setup-python@v4
         with:
-          python-version: 3.8
+          python-version: '3.10'
 
       - name: Install dependencies
         run: |

README.md

@@ -26,8 +26,9 @@ everything that I'll talk about is covered in more detail somewhere else (even
 if that somewhere is just a comment in some source code), but hopefully this
 summary can point you in the right direction if you have a use case like this.
 
-**A warning**: I'm writing this in January 2021 and much of what I'm talking
-about is based on essentially undocumented APIs that are likely to change.
+**A warning**: I'm writing this in January 2021 (most recent update November 2023; see
+github for the full revision history) and much of what I'm talking about is based on
+essentially undocumented APIs that are likely to change.
 Furthermore, I'm not affiliated with the JAX project and I'm far from an expert
 so I'm sure there are wrong things that I say. I'll try to update this if I
 notice things changing or if I learn of issues, but no promises! So, MIT license
@@ -358,7 +359,7 @@ from jax.lib import xla_client
 from kepler_jax import cpu_ops
 
 for _name, _value in cpu_ops.registrations().items():
-    xla_client.register_cpu_custom_call_target(_name, _value)
+    xla_client.register_custom_call_target(_name, _value, platform="cpu")
 ```
 
 Then, the **lowering rule** is defined roughly as follows (the one you'll
@@ -400,13 +401,13 @@ def _kepler_lowering(ctx, mean_anom, ecc):
     return custom_call(
         op_name,
         # Output types
-        out_types=[dtype, dtype],
+        result_types=[dtype, dtype],
         # The inputs:
         operands=[mlir.ir_constant(size), mean_anom, ecc],
         # Layout specification:
         operand_layouts=[(), layout, layout],
         result_layouts=[layout, layout]
-        )
+        ).results
 
 mlir.register_lowering(
         _kepler_prim,
@@ -651,15 +652,15 @@ def _kepler_lowering_gpu(ctx, mean_anom, ecc):
     return custom_call(
         op_name,
         # Output types
-        out_types=[dtype, dtype],
+        result_types=[dtype, dtype],
         # The inputs:
         operands=[mean_anom, ecc],
         # Layout specification:
         operand_layouts=[layout, layout],
         result_layouts=[layout, layout],
         # GPU-specific additional data for the kernel
         backend_config=opaque
-    )
+    ).results
 
 mlir.register_lowering(
         _kepler_prim,

setup.py

@@ -120,7 +120,10 @@ def build_extension(self, ext):
     packages=find_packages("src"),
     package_dir={"": "src"},
     include_package_data=True,
-    install_requires=["jax", "jaxlib"],
+    install_requires=[
+        "jax>=0.4.16",
+        "jaxlib>=0.4.16"
+    ],
     extras_require={"test": "pytest"},
     ext_modules=extensions,
     cmdclass={"build_ext": CMakeBuildExt},

src/kepler_jax/kepler_jax.py

@@ -7,7 +7,7 @@
 import numpy as np
 from jax import core, dtypes, lax
 from jax import numpy as jnp
-from jax.abstract_arrays import ShapedArray
+from jax.core import ShapedArray
 from jax.interpreters import ad, batching, mlir, xla
 from jax.lib import xla_client
 from jaxlib.hlo_helpers import custom_call
@@ -16,7 +16,7 @@
 from . import cpu_ops
 
 for _name, _value in cpu_ops.registrations().items():
-    xla_client.register_cpu_custom_call_target(_name, _value)
+    xla_client.register_custom_call_target(_name, _value, platform="cpu")
 
 # If the GPU version exists, also register those
 try:
@@ -93,13 +93,13 @@ def _kepler_lowering(ctx, mean_anom, ecc, *, platform="cpu"):
         return custom_call(
             op_name,
             # Output types
-            out_types=[dtype, dtype],
+            result_types=[dtype, dtype],
             # The inputs:
             operands=[mlir.ir_constant(size), mean_anom, ecc],
             # Layout specification:
             operand_layouts=[(), layout, layout],
             result_layouts=[layout, layout]
-        )
+        ).results
 
     elif platform == "gpu":
         if gpu_ops is None:
@@ -113,15 +113,15 @@ def _kepler_lowering(ctx, mean_anom, ecc, *, platform="cpu"):
         return custom_call(
             op_name,
             # Output types
-            out_types=[dtype, dtype],
+            result_types=[dtype, dtype],
             # The inputs:
             operands=[mean_anom, ecc],
             # Layout specification:
             operand_layouts=[layout, layout],
             result_layouts=[layout, layout],
             # GPU specific additional data
             backend_config=opaque
-        )
+        ).results
 
     raise ValueError(
         "Unsupported platform; this must be either 'cpu' or 'gpu'"