search

smousavi05 / EQTransformer

EQTransformer, a python package for earthquake signal detection and phase picking using AI.
https://rebrand.ly/EQT-documentations
MIT License
301 stars 148 forks source link

EQTransformer training parameters #111

Closed alirezaniki closed 2 years ago

alirezaniki commented 2 years ago

Dear Mostafa,

I am going to train EQT on a new dataset in a way in which at last I can compare it with EQT. I've read the paper but I am still not sure about the values for cnn_blocks and lstm_blocks in the test_trainer code. Am I supposed to use the default values? I mean, are those values the ones that EQT has been trained on?

Many thanks, Alireza.

smousavi05 commented 2 years ago

Hi @alirezaniki depending on your dataset, the optimal values for the architecture might change. You can use hyperparameter tuning packages to find the optimal values for your case. The default values are those that worked best for STEAD dataset. BTW, you can use those packages like seibench for converting your data into STEAD format and retraining EqT.

alirezaniki commented 2 years ago

@smousavi05 Thank you so much for your comments. Actually, my new dataset is derived from STEAD, so I'll try to use and fine-tune the default parameters.