search

awslabs / gluonts

Probabilistic time series modeling in Python
https://ts.gluon.ai
Apache License 2.0
4.47k stars 744 forks source link

LSTNet's scaling #765

Closed kashif closed 4 years ago

kashif commented 4 years ago

Description

The scaling in LSTNet is only valid for NTC tensors however we are feeding it NCT tensors since output_NTC=False and thus are not using the scaled values.

To Reproduce

(Please provide minimal example of code snippet that reproduces the error. For existing examples, please provide link.)

put code here

Error message or code output

(Paste the complete error message, including stack trace, or the undesired output that the above snippet produces.)

put error or undesired output here

Environment

(Add as much information about your environment as possible, e.g. dependencies versions.)

lostella commented 4 years ago

Thanks @kashif, you’re right. I think this requires having an axis option in scaler blocks, what do you think?

kashif commented 4 years ago

@lostella right I have that working, also the the LSTNet implementation is wrong, and so I can send both the fixes in 1 PR... is that ok?

lostella commented 4 years ago

@lostella right I have that working, also the the LSTNet implementation is wrong, and so I can send both the fixes in 1 PR... is that ok?

Hey thanks for that, it would be best to have two separate PRs I think

kashif commented 4 years ago

ok PR with axis coming first

kashif commented 4 years ago

fixed