Tom Andersson comments

Results 17 comments of


                                            Tom Andersson

Antarctica data generator always loads PyTorch

I suspect this can be closed as this is no longer used? Thoughts @stratisMarkou?

Antarctica data generator always loads PyTorch

Okay my bad, no probs :-)

[FR] Ability to draw cheap AR samples with subset-condition-predict procedure with `nps.ar_predict`

P.S. It would also be great to also expose the `N / K` cost reduction approach described in the 'Consistency and the AR design space' paragraph.

[FR] Ability to draw cheap AR samples with subset-condition-predict procedure with `nps.ar_predict`

This [subset-condition-predict feature is now supported in DeepSensor](https://tom-andersson.github.io/deepsensor/user-guide/prediction.html#advanced-autoregressive-sampling) (my package that wraps `neuralprocesses` and adds functionality for environmental sciences). It may still be useful to support directly in `neuralprocesses` however!

[FR] Allow `nps.loglik`, `nps.ar_loglik` and `nps.ar_predict` to work with numpy arrays

I think this would do the job nicely @wesselb! Let me know when you've pushed it and I can update `backends` and test it out.

[FR] Allow `nps.loglik`, `nps.ar_loglik` and `nps.ar_predict` to work with numpy arrays

I don't think more is needed than the `tf.cast` you propose - this should fix the dtype errors when running the `nps.loglik` (etc) methods. The only issue I can foresee...

[Feature]: Hyperparameter sliders for logged media

Hi Leslie, yes my request can be framed in a very generic way. For sweeps that log charts/media, I would like to have a sweep chart that shows how that...