Charles Chen

Results 7 comments of Charles Chen

Dataflow jobs fail when using the same beam_pipeline_args for InteractiveContext and KubeflowDagRunner

Thanks! Could you check whether changing the following line fixes this? Specifically, change this line: https://github.com/tensorflow/tfx/blob/6429c643233f1c1fca41c7c02e7da966c763c7eb/tfx/orchestration/experimental/interactive/interactive_context.py#L139 to ```python beam_pipeline_args = list(beam_pipeline_args or self.beam_pipeline_args) ```

`LocalDagRunner` ignores output artifacts from container components

Hi @IanTayler, thank you for the report and apologies for the late reply. Currently, you are right that the current implementation does not use volumes to achieve value-passing here. In...

udf_utils: fix python path modification

Thank you for this change! This looks good to me.

udf_utils: fix python path modification

It looks like I don't have permissions to authorize merges anymore. @jiyongjung0, could you ping me with instructions on the new process?

Ability to name dataflow jobs

Thanks, I will look into adding this functionality.

How to 'bring into TFX flow' a custom pkl file generated 'outside of TFX'

Hi @cyrilthank, we do not have plans to add support for this in the near future. CC: @zhitaoli @ruoyu90

How to 'bring into TFX flow' a custom pkl file generated 'outside of TFX'

Hey @cyrilthank, you can look into writing your own component for adding the functionality you need. You can see examples of how to do this with the standard TFX components...