I'm running large-ish models with mxTryHard, but notice checkpoints aren't an option as with mxRun. Is there still a way to save / load progress as I go?
I was hoping that mxTryHardOrig, which has a ... argument, would pass checkpoint to mxRun. It doesn't. About the only way I see around it at this point is to do mxTryHard manually, using omxSetParameters() to put in random starting values, and to monitor the output. There is a bit of a snag, however, that the model name is used to label the checkpoint file, so it would likely get overwritten with each fitting attempt, and then not match the path towards the solution. Probably such considerations are why it's not implemented at all. That and the day being only 24hrs long etc...
I think checkpointing during mxTryHard() could be implemented so that the new checkpointing output is appended to the output file, rather than overwriting it.
Also, mxJiggle() is a useful function for randomly perturbing start values manually.
I was hoping that mxTryHardOrig, which has a ... argument, would pass checkpoint to mxRun. It doesn't. About the only way I see around it at this point is to do mxTryHard manually, using omxSetParameters() to put in random starting values, and to monitor the output. There is a bit of a snag, however, that the model name is used to label the checkpoint file, so it would likely get overwritten with each fitting attempt, and then not match the path towards the solution. Probably such considerations are why it's not implemented at all. That and the day being only 24hrs long etc...
I was hoping the same thing. Just wanted to make sure I wasn't missing anything before I jumped into the TryHard code... :)
I'll put it on the wishlist.
I think checkpointing during
mxTryHard()
could be implemented so that the new checkpointing output is appended to the output file, rather than overwriting it.Also,
mxJiggle()
is a useful function for randomly perturbing start values manually.https://openmx.ssri.psu.edu/node/4643