Closed SuryanarayanaY closed 9 months ago
All modified lines are covered by tests :white_check_mark:
Comparison is base (
3a464f9
) 83.75% compared to head (567dbaa
) 60.56%.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
APologies. PR created by mistake here as source code link points it to here.
Created a fresh PR in keras-team/keras with PR No. #18599 and hence closing here.
Thanks!
At present,the API keras_core.input documentation states that the argument dtype to be passed as a string such as 'tf.float32' etc. However if we pass
tf.dtypes.DType
such astf.float32
,tf.float64
or atorch.dtype
such astorch.float32
,torch.float64
it will works as well as the internal code calls keras_core.backend.standardize_dtype which can converttf.float32
ortorch.float32
to string such as 'float32'.IMO, its better to mention this is documentation that this API can also handle both
tf.dtypes.DType
andtorch.dtype
so that users are aware of this and they are free to choose any of the above ways to pass value to dtype.Attached gist as reference for the above exercise.