tff.jax.computation
Stay organized with collections
Save and categorize content based on your preferences.
Decorates/wraps Python functions containing JAX code as TFF computations.
tff.jax.computation(
*args, **kwargs
)
This wrapper can be used in a similar manner to tff.tensorflow.computation,
with exception of the following:
The code in the wrapped Python function must be JAX code that can be
compiled to XLA (e.g., code that one would expect to be able to annotate
with @jax.jit).
The inputs and outputs must be tensors, or (possibly recursively) nested
structures of tensors. Sequences are currently not supported.
Example:
@tff.jax.computation(np.int32)
def comp(x):
return jax.numpy.add(x, np.int32(10))
Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.
Last updated 2024-04-26 UTC.
[null,null,["Last updated 2024-04-26 UTC."],[],[],null,["# tff.jax.computation\n\n\u003cbr /\u003e\n\nDecorates/wraps Python functions containing JAX code as TFF computations. \n\n tff.jax.computation(\n *args, **kwargs\n )\n\nThis wrapper can be used in a similar manner to [`tff.tensorflow.computation`](../../tff/tensorflow/computation),\nwith exception of the following:\n\n- The code in the wrapped Python function must be JAX code that can be\n compiled to XLA (e.g., code that one would expect to be able to annotate\n with `@jax.jit`).\n\n- The inputs and outputs must be tensors, or (possibly recursively) nested\n structures of tensors. Sequences are currently not supported.\n\n#### Example:\n\n @tff.jax.computation(np.int32)\n def comp(x):\n return jax.numpy.add(x, np.int32(10))"]]
