search

thouska / spotpy

A Statistical Parameter Optimization Tool
https://spotpy.readthedocs.io/en/latest/
MIT License
247 stars 149 forks source link

Can't run throught the getting started example #291

Open qiyang77 opened 1 year ago

qiyang77 commented 1 year ago

https://spotpy.readthedocs.io/en/latest/getting_started/ Seems like the type of the "results" variable is inconsistent between results=sampler.getdata() and spotpy.analyser.plot_parameterInteraction(results)

thouska commented 1 year ago

Hi @qiyang77, thank you for your message. I just tested the getting_started.py script and did not receive any errors. Could you explain what you mean with "variable is inconsistent"? The results are typically numpy.ndarray within spotpy which can be read/used/plotted in the spotpy.analyser.

qiyang77 commented 1 year ago

Oh I see. It is because the "pip install spotpy" will not install the last version so the tutorial can't be run through. After I installed spotpy manually everything is fine. One issue is that my model is a pytorch-based neural network, and I used the SCE-UA to calibrate seven model parameters, however, the iteration process is very slow. Can we accelerate the neural network inference via batch mode?

thouska commented 1 year ago

Certainly possible. However, there is currently no development on spotpy in that direction. However, feel free to support us with a pull request. Although I am not sure what you mean with "very slow". Usually, spotpy consumes very little time to produce repetitions, the bottleneck is in the majority of the cases the applied model.