Backend Guide

BaseAttentive uses Keras 3 as its neural network runtime. Keras 3 supports three backends — TensorFlow, Torch, and JAX — and BaseAttentive inherits that flexibility.

Backends

Support Status

Backend

Status

Notes

TensorFlow

Supported

Full BaseAttentive model path; tested on Linux, macOS, Windows.

Torch

Supported

Full model path tested; includes CPU, CUDA, and Apple MPS devices.

JAX

Supported

Runtime abstraction available; full model path tested on CPU/TPU.

Selecting a Backend

Set the backend before importing base_attentive or keras:

export KERAS_BACKEND=tensorflow   # or torch / jax
export BASE_ATTENTIVE_BACKEND=tensorflow

Or in Python:

import os
os.environ["KERAS_BACKEND"] = "torch"   # set before any keras import
from base_attentive import BaseAttentive

Runtime inspection and control:

from base_attentive import (
    get_backend,
    get_available_backends,
    get_backend_capabilities,
    set_backend,
    normalize_backend_name,
)

backend = get_backend()
print(get_available_backends())
print(get_backend_capabilities("jax"))
print(normalize_backend_name("tf"))   # -> "tensorflow"

How Backend Resolution Works

When no explicit backend is specified, BaseAttentive checks in this order:

  1. BASE_ATTENTIVE_BACKEND environment variable

  2. KERAS_BACKEND environment variable

  3. The backend previously set in the current Python process via set_backend()

  4. The best available backend detected automatically

  5. tensorflow as the final fallback

Detection and Selection Utilities

from base_attentive import (
    detect_available_backends,
    select_best_backend,
    ensure_default_backend,
)

# Inspect all installed backends
info = detect_available_backends()
for name, details in info.items():
    print(f"{name}: available={details.get('available')}, "
          f"version={details.get('version')}")

# Select the best available backend
best = select_best_backend(require_supported=True)
print(f"Selected backend: {best}")

# Ensure a default backend is ready
name = ensure_default_backend(auto_install=False)

Version Compatibility

from base_attentive.backend import (
    check_tensorflow_compatibility,
    check_torch_compatibility,
    get_backend_version,
    version_at_least,
)

ok, msg = check_tensorflow_compatibility()
print(msg)   # e.g. "TensorFlow 2.15.0 is compatible"

ok, msg = check_torch_compatibility()

ver = get_backend_version("tensorflow")
ok = version_at_least("2.13.0", "2.12.0")   # True

See Also