Open anandsahuofficial opened 2 days ago
Hi @anandsahuofficial . The library is undergoing a lot of changes at moment, but I've made a release associated to the paper. Can you make sure you have installed the 0.1 version of the library?
Hi Rizzi, 0.1 version of the library is working fine. I ran run_tfep.py for just one molecule and it is taking ages with 28 Cores.
I ran _"python runtfep.py --molid=00140610 --flow=zmatrix-spline --batch=48 --subdir=zmatrix-spline-b48" with 28 Cores for a day and I am still on 646/20834 progress "INFO:modules.trainer:Starting epoch 1/1, batch 646/20834".
How much time does it take to train the network and check the analysis to reproduce the figures in Manuscript?
The driver script in the paper-multimap-tfep-2023
repo was written to work on SLURM. You may need to change the parallelization configuration to make it exploit better the resources (see for example here).
Also, we ran with batch size 48 because we had 48 CPU cores over which to parallelize it. Try with batch size 28 (or a number that is divisible by 28) or you'll lose a lot of performance.
On our cluster, each calculation took a few hours (there are timings in the supporting information). That depends a lot on the computing architecture though.
Hi I was going through your 2023 paper on Multimap-TEFP and tried installing and running the material your provided through in the repo "paper-multimap-tfep-2023". I am facing this particular AttributeError issue. I used the installation direction of tfep as instructed. Do you mind solving this issue for me?
Thanks, Anand