search

liufeng2317 / ADsurf

An automatic differentiation based (AD-based) multimodal surface wave inversion tools
MIT License
33 stars 3 forks source link

Issue with MonteCarlo Search for initial model #2

Open Akumarres opened 10 months ago

Akumarres commented 10 months ago

Dear Liu,

I am also facing the issue while running the 01_3-layer-sandwitch-Multimodal.ipynb jypyter file.

please check the attached files. Screenshot 2024-01-03 at 11-10-46 01_3-layer-sandwitch-Multimodal - Jupyter Notebook Screenshot 2024-01-03 at 11-10-13 01_3-layer-sandwitch-Multimodal - Jupyter Notebook

I need your help to resolve the issue. Thank you

liufeng2317 commented 10 months ago

Hello This error is due to inconsistent inversion of tensor. I forced all tensors to be float32. You can try to update the code and run it again.

Akumarres commented 10 months ago

hello, After update I am getting new error while running inversion.

Screenshot 2024-01-03 at 14-55-02 01_3-layer-sandwitch-Multimodal - Jupyter Notebook

Thank you

liufeng2317 commented 10 months ago

this means that you not have cuda to run the code, you can change the device to "cpu" e.g. device = "cpu"

Akumarres commented 10 months ago

I have uses the command given below as it was recommended in README.md

$ conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c cpu

Also , all requirements are install. some of them are upgraded version.

I took time to recheck everything. I can share the list also if you needed.

Thank you.

Akumarres commented 10 months ago

One more Information I would like to share with you. While installing the requirement given below.

torch==1.13.1+cu116 torchaudio==0.13.1+cu116 torchvision==0.14.1+cu116

Because of "+cu116" it was showing some error message. ERROR: Could not find a version that satisfies the requirement torch==1.13.1+cu116 etc.

Thank you

liufeng2317 commented 7 months ago

Dear Akumarres: Apologies for the delayed response. Based on my local testing, I'm unsure if torch version 1.13.1 can support this library. If it does, you might want to consider trying it with a newer version of torch; torch 2.0.1 has been verified to work. Rest assured, I'll update the program to accommodate a wider range of torch versions.