search

microsoft / causica

MIT License
442 stars 55 forks source link

Run time using GPU is vey slow #96

Open kh2014 opened 9 months ago

kh2014 commented 9 months ago

Hello

I used a Dataset which is a CSV file containing 58 columns and 50000 raws I used 16GO GPU, It' took more than 1 hour to get the causal graph.

WenboGong commented 9 months ago

Does the 1 hour include the training time for DECI?

If so, it is expected because DECI is a deep learning based method, and need to iterate through your dataset with mini-batch. If the batch size is small, and number of epochs is large, it will take long time to train the model.

So one suggestion is to use slightly larger batch size and monitor the training progress to decide a suitable training epochs.

kh2014 commented 9 months ago

Thank you for your response. One hour is just for training. Do you have an example showing how to use mini-batches? I would be grateful if you could also provide an example of using DECI with custom data.