Missing npm package for v0.47.1

[dante-lee]

Details

• There seems no npm package for tensorflow_model_analysis==0.47.1(link)
• Currently we make HTML file of evaluation result with ipywidgets package(see below code)
• If we run below code with tensorflow_model_analysis==0.47.1 and then open that html with browser(using Chrome & Safari), we get error something like this

  cannot find https://unpkg.com/tensorflow_model_analysis@0.47.1/dist/index.js

• I think it is caused by missing npm package for 0.47.1
• Can you update npm package for 0.47.1(Cannot use 0.47.0 or older version since we use TFX 1.16 and it requires tensorflow_model_analysis >= 0.47.0 and 0.47.0 has circular import error)

Code for html

import tensorflow_model_analysis as tfma
from ipywidgets.embed import embed_minimal_html

slicing_specs = ...
eval_result = tfma.load_eval_result(...)
metric_views = []
for slicing_spec in slicing_specs:
    metric_views.append(tfma.view.render_slicing_metrics(eval_result, slicing_spec=slicing_spec))

embed_minimal_html("result.html", metric_views)

[janasangeetha]

Hey @dante-lee
Sorry for addressing the issue late. Thank you for bringing this to our attention. We will work on this and keep you updated.
Thank you!

[dante-lee]

@janasangeetha
Thank you for your consideration on this issue. Is there any update on this issue?

[janasangeetha]

Hey @dante-lee
We have released latest version 0.48.0. Please try the same and let us know if you are still facing any issues.
Thank you!

[dante-lee]

@janasangeetha
Thanks for your reply!

I have one more question. I'm currently using TFX 1.16.0 and according to dependencies.py in tfx source code, it seems like tfx==1.16.0 requires tensorflow-model-analysis 0.47.x which could lead to version conflict.

I wonder is it ok to use tensorflow-model-analysis 0.48.0 with TFX 1.16.0.(On my local machine, I was failed to install tfx 1.16 with tensorflow-model-analysis 0.48.0)

[janasangeetha]

Hey @dante-lee
We are working on upgrading TFX. We’ll keep you updated as soon as we have more information.
Appreciate your patience!

[dante-lee]

@janasangeetha
OK thanks!