tf.compat.v1.to_bfloat16

Casts a tensor to type bfloat16. (deprecated)

Migrate to TF2

This name was deprecated and removed in TF2, but has an exact replacement tf.cast(..., tf.bfloat16). There are no further issues with eager execution or tf.function.

Before:

tf.compat.v1.to_bfloat16(tf.constant(3.14, dtype=tf.float32))
<tf.Tensor: shape=(), dtype=bfloat16, numpy=3.14>

After:

tf.cast(tf.constant(3.14, dtype=tf.float32), tf.bfloat16)
<tf.Tensor: shape=(), dtype=bfloat16, numpy=3.14>