openclimatefix power_perceiver issues

openclimatefix / power_perceiver

Machine learning experiments using the Perceiver IO model to forecast the electricity system (starting with solar)

MIT License

7 stars 1 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

FIX MAJOR BUG IN GSP ALIGNMENT

#200 JackKelly opened 1 year ago
0
Remove setting GSP to One option once retrained

#199 jacobbieker opened 2 years ago
0
Change to using OCF Datapipe's BatchKey

#198 jacobbieker closed 2 years ago
1
[pre-commit.ci] pre-commit autoupdate

#197 pre-commit-ci[bot] opened 2 years ago
1
[pre-commit.ci] pre-commit autoupdate

#196 pre-commit-ci[bot] closed 2 years ago
0
Prepared Batches after processing seem to have differing numbers of timesteps, either 32 or 31

#195 jacobbieker closed 2 years ago
2
Topographic Data isn't appended correctly with Batch

#194 jacobbieker closed 2 years ago
1
[pre-commit.ci] pre-commit autoupdate

#193 pre-commit-ci[bot] closed 2 years ago
1
Update Preprepared Dataset to current v15_v2 dataset

#192 jacobbieker closed 2 years ago
2
add fastai fix

#191 peterdudfield closed 2 years ago
1
Power Perceiver Refactor

#190 jacobbieker closed 2 years ago
3
Pydantic model for batch --> power perceiver

#189 peterdudfield closed 2 years ago
2
Make pydantic object ready for PP model

#188 peterdudfield closed 2 years ago
0
Fix current tests so that work on CI

#187 peterdudfield opened 2 years ago
0
Issue/182 GitHub actions

#186 peterdudfield closed 2 years ago
0
Test: PP on Batch

#185 peterdudfield closed 2 years ago
1
Convert 'Batch' to model dict

#184 peterdudfield closed 2 years ago
1
Move experiment 27 in production folder

#183 peterdudfield closed 2 years ago
0
Add github actions to run tests

#182 peterdudfield closed 2 years ago
0
[pre-commit.ci] pre-commit autoupdate

#181 pre-commit-ci[bot] closed 2 years ago
0
Automatically detect "bad" PV data

#180 JackKelly opened 2 years ago
3
Use Transformer-XL approach to attend to many hours of recent history

#179 JackKelly opened 2 years ago
0
Load `BatchML` into the `power_perceiver.load_prepared_batches` code

#177 JackKelly opened 2 years ago
2
train on more data. Enable the data loaders to take a list of periods to load, so we can validate on, say, every 5th week of 2021.

#176 JackKelly closed 2 years ago
1
Predict multiple GSPs at once

#175 JackKelly opened 2 years ago
0
Recall similar days from the history

#174 JackKelly opened 2 years ago
7
More heads & higher D_MODEL

#173 JackKelly opened 2 years ago
0
More Gaussians

#172 JackKelly closed 2 years ago
1
Tweak plot of national pv: Include date & y-axis label & title.

#171 JackKelly closed 2 years ago
0
Use hydra to specify config options

#170 JackKelly opened 2 years ago
0
Move all the models into their own python files within the main library

#169 JackKelly closed 2 years ago
1
Increase history

#168 JackKelly opened 2 years ago
0
Write prepared batches to disk as a separate script

#167 JackKelly opened 2 years ago
0
Use the same order of features for each query.

#166 JackKelly closed 2 years ago
2
NWP query needs solar position, and step.

#165 JackKelly closed 2 years ago
2
Plot NWPs

#164 JackKelly closed 2 years ago
1
Try giving the NWP as multiple query elements. e.g. each variable is its own element (with a learnt channel ID)

#163 JackKelly closed 2 years ago
2
Don't reshape_time_as_batch in `GSPQueryGenerator`!

#162 JackKelly opened 2 years ago
0
Check that we're correctly masking whole examples when they have no PV

#161 JackKelly closed 2 years ago
1
Write unittest to check that `SunPosition` propogates NaNs when there are no PV systems in example

#160 JackKelly opened 2 years ago
0
Why do we sometimes get NaNs in the PV data?

#159 JackKelly closed 2 years ago
3
Modify `GSPQueryGenerator` so it can create both 5-minutely GSP queries, and half-hourly GSP queries

#158 JackKelly closed 2 years ago
1
Modify the experiment 27 code so it looks more like experiment 29 code. In particular: remove all the ugly reshaping from the main model, and let the `satellite_processor` etc. handle the reshaping.

#157 JackKelly closed 2 years ago
0
Continue training the U-Net

#156 JackKelly opened 2 years ago
0
Try Flamingo / contrastive learning approach to solar PV forecasting

#155 JackKelly opened 2 years ago
10
Try Perceiver again. With tweaks from the Flamingo paper

#154 JackKelly opened 2 years ago
0
Get experiment 27 working again with the latest code.

#153 JackKelly closed 2 years ago
0
Push model weights to HuggingFace.

#152 JackKelly closed 2 years ago
2
Create minimal `requirements.txt` for inference in production

#151 JackKelly closed 2 years ago
0
Branch model 27.01

#150 JackKelly closed 2 years ago
1