search

openclimatefix / predict_pv_yield

Using optical flow & machine learning to predict PV yield
MIT License
52 stars 11 forks source link

huggingface model name for OCF OptiFlow model #113

Open davidburtcam opened 1 year ago

davidburtcam commented 1 year ago

Really interesting results and great project!

I was going through the code, the March'22 blog post and the research report. I would like to reproduce the results of "OCF OptiFlow model".

Can you please let me know the model name for "OCF optiFlow model" in the huggingface models ?

Also, wanted to confirm if that is the best performing model so far.

Thank you

JackKelly commented 1 year ago

Hi David, thanks for getting in touch, and sorry for the very slow reply!

I'm afraid we're not actively developing OptiFlow any more, and its weights aren't on HF.

We are developing some new models at the moment. We hope to share the code & weights for these new models as soon as they are ready (but that might be months away, I'm afraid)

tomasvanoyen commented 1 year ago

Update:

considering the context described in nowcasting, it appears that the OptiFlow results should come from predict_pv_yield while satflow should be something better as there the movement of clouds is predicted with more advanced tricks and not just optical flow techniques.

@JackKelly, am I correct? Perhaps it could be worthwhile to mention the context in the readme as it is otherwise quite difficult to follow.

Regards,

Tomas

Hi @JackKelly ,

I realize that OptiFlow is not in active development still. Yet, it would be nice to know which code (repo) to use to reproduce the results as mentioned in the report.

Should I start off with predict_pv_yield or satflow or both?

Looking forward.

Regards,

Tomas

ps: @davidburtcam if in the meantime you found the solution, happy to hear also from you.