-
Fixed the bug appeared in the process of installing the boost library.
-
Fixed the bug in the saturation operator that caused possible failure in the gradient test. This behavior appeared when certain compilers were used, and it might lead to some convergence issues. Now the saturation operator passes the gradient test consistently.
-
Changed the legacy name PyTensorFlow to ADCME in CustomOps.
-
TODO:
- In /docs/codes, we host some legacy codes. Need to change the name from PyTensorFlow to ADCME in those scripts, and make necessary modifications to load custom ops, so that the results in the WRR paper can be reproduced using ADCME.
- Add more tests in addition to TestFWI.jl