(<code>e</code> or <code>E</code> followed by an optional sign and decimal digits).
One of the integer part or the fractional part may be elided; one of the decimal point
or the exponent part may be elided.
An exponent value exp scales the mantissa (integer and fractional part) by 10<sup>exp</sup>.
</p>

<p>
A hexadecimal floating-point literal consists of a <code>0x</code> or <code>0X</code>

* <IF A CHANGE CLOSES A GITHUB ISSUE, IT SHOULD BE DOCUMENTED HERE>
* <NOTES SHOULD BE GROUPED PER AREA>

* `tf.lite`
    * Quantization for `FullyConnected` layer is switched from per-tensor to
      per-channel scales for dynamic range quantization use case (`float32`
      inputs / outputs and `int8` weights). The change enables new quantization
      schema globally in the converter and inference engine. The new behaviour