add automatic backend detection and selection

Author: LarsKue

bayesflow/__init__.py

@@ -1,27 +1,62 @@
-from . import (
-    approximators,
-    adapters,
-    augmentations,
-    datasets,
-    diagnostics,
-    distributions,
-    experimental,
-    networks,
-    simulators,
-    utils,
-    workflows,
-    wrappers,
-)
-
-from .adapters import Adapter
-from .approximators import ContinuousApproximator, PointApproximator
-from .datasets import OfflineDataset, OnlineDataset, DiskDataset
-from .simulators import make_simulator
-from .workflows import BasicWorkflow
+# ruff: noqa: E402
+# disable E402 to allow for setup code before importing any internals (which could import keras)
 
 
 def setup():
     # perform any necessary setup without polluting the namespace
+    import os
+    from importlib.util import find_spec
+
+    issue_url = "https://github.com/bayesflow-org/bayesflow/issues/new?template=bug_report.md"
+
+    if "KERAS_BACKEND" not in os.environ:
+        # check for available backends and automatically set the KERAS_BACKEND env variable or raise an error
+        class Backend:
+            def __init__(self, display_name, package_name, env_name, install_url):
+                self.display_name = display_name
+                self.package_name = package_name
+                self.env_name = env_name
+                self.install_url = install_url
+
+        backends = [
+            Backend("JAX", "jax", "jax", "https://docs.jax.dev/en/latest/quickstart.html#installation"),
+            Backend("PyTorch", "torch", "torch", "https://pytorch.org/get-started/locally/"),
+            Backend("TensorFlow", "tensorflow", "tensorflow", "https://www.tensorflow.org/install"),
+        ]
+
+        found_backends = []
+        for backend in backends:
+            if find_spec(backend.package_name) is not None:
+                found_backends.append(backend)
+
+        if not found_backends:
+            message = "No suitable backend found. Please install one of the following:\n"
+            for backend in backends:
+                message += f"{backend.display_name}\n"
+            message += "\n"
+
+            message += f"If you continue to see this error, please file an bug report at {issue_url}.\n"
+            message += (
+                "You can manually select a backend by setting the KERAS_BACKEND environment variable as shown below:\n"
+            )
+            message += "https://keras.io/getting_started/#configuring-your-backend"
+
+            raise ImportError(message)
+        elif len(found_backends) > 1:
+            message = "Multiple backends found:\n"
+            for backend in found_backends:
+                message += f"- {backend.display_name}\n"
+            message += "\n"
+
+            message += (
+                "You can manually select a backend by setting the KERAS_BACKEND environment variable as shown below:\n"
+            )
+            message += "https://keras.io/getting_started/#configuring-your-backend"
+
+            raise ImportError(message)
+        else:
+            os.environ["KERAS_BACKEND"] = found_backends[0].env_name
+
     import keras
     import logging
 
@@ -60,3 +95,24 @@ def setup():
 # call and clean up namespace
 setup()
 del setup
+
+from . import (
+    approximators,
+    adapters,
+    augmentations,
+    datasets,
+    diagnostics,
+    distributions,
+    experimental,
+    networks,
+    simulators,
+    utils,
+    workflows,
+    wrappers,
+)
+
+from .adapters import Adapter
+from .approximators import ContinuousApproximator, PointApproximator
+from .datasets import OfflineDataset, OnlineDataset, DiskDataset
+from .simulators import make_simulator
+from .workflows import BasicWorkflow