Update README.md: Ensembles #59
Conversation
Updated to reflect the newly added ensemble capability, wherein the output of the LSTM module is the ensemble mean forecast of the output from multiple trained LSTMs. I made relevant changes through the "Trained LSTM Model" section, but the below sections on Dependencies thru Unit Test I did not change (yet), but I think that much of it will be unchanged. I had a few questions though: I think we are using AORC forcings, but I do not see those stored in the "Data" folder despite the data section of the readme saying the necessary data lives there. I am not sure where the AORC data lives (in Jonathan's fork of LSTM which is more updated it appears to come from another repo entirely?). Please advise on where I can point a user to the relevant forcings data.
|
Hi @Zacmce , the PR I made does include updated readme with changes for the examples, including instructions on accessing the example AORC data from this link. Below is a link to my updated readme, and below that is a screen shot of the changed lines. But note that @aaraney also has a PR to do the ensembling in the different manner we talked about last week. https://github.com/jmframe/lstm/blob/c3a93c8368b9adcb695dd07e62484436aee4cd9a/README.md?plain=1#L17
|
|
@jmframe ah I see. Thanks for bringing to my attention! Austin asked if I could do a readme update but perhaps he was referring to his PR not yours... And the documentation I added was outlining the implementation that we did the other day reflected in your PR. Don't a few other sections need to be updated in the readme though to reflect the ensemble capability and overall current implementation? Like the Trained LSTM Models section, where we no longer have _all_attributes etc. |
|
Im sorry @jmframe and @Zacmce! I didn't communicate to @jmframe that i'd asked @Zacmce to make a PR to update the readme. To @Zacmce's point, can you both collaborate on this PR to update the readme as #56 will not be merged? For clarity, #56 will not be merged but is still a great place for conversation. All the content (code, etc.) in #56 is still included in #57. Meaning the commits from #56 have been condensed into a single commit and then overwritten by my contributions but the history is still tracked by |
Yes, absolutely right. I've updated my README now to reflect those trained models: |
Updated to reflect the newly added ensemble capability, wherein the output of the LSTM module is the ensemble mean forecast of the output from multiple trained LSTMs. I made relevant changes through the "Trained LSTM Model" section, but the below sections on Dependencies thru Unit Test I did not change (yet), but I think that much of it will be unchanged. I had a few questions though:
I think we are using AORC forcings, but I do not see those stored in the "Data" folder despite the data section of the readme saying the necessary data lives there. I am not sure where the AORC data lives (in Jonathan's fork of LSTM which is more updated it appears to come from another repo entirely?). Please advise on where I can point a user to the relevant forcings data.
Once the final PR is completed to merge jmframe/lstm onto owp/lstm, I will want to do another update of the Readme to reflect the latest links so I can point the reader to e.g. the most recent notebooks and files, that do not currently exist on owp/lstm because the merge has not been fully completed yet.
Checklist
Testing checklist
Target Environment support
Accessibility
Other