Small problems running the script and information

[pablo-arantes]

Small problems running the script and information

[pablo-arantes]

Hi @ggoetten

Thank you for your message! I'm glad our pipeline is helping you. Our main goal when we started to develop it was to give computational power for everyone (Make it rain!)... :-)

Regarding your questions, follow the answers:

1. Unfortunately, it is not possible to continue this last step where it has stopped. That was the reason we have been using strides, the user can calculate how many nanoseconds per day and then select the number of strides and simulation time for 12 hours of run (free) or 24 hours (pro). I understand that sometimes Google force to stop and unfortunately, we should run the last step from the beginning. My suggestion is to create a higher number of strides with a shorter simulation time. Thereby, if you lose your last stride because of Google interruptions, it will be fast to run again.

2. Yes, you should run the installation of dependencies every time you log in Colab. When you connect a GPU from Google, you are creating a virtual machine. So, every time you need to install the dependencies again. At least, the installation takes no more than 7 minutes. This is very common on Colab, if you are going to run a different script for example, with machine learning stuffs, you should install the libraries every time you connect too.

3. Our benchmark was made with Amber force field, not only that, depends on the size of the box. Maybe you are using a bigger box than us, with more waters and so on.

I hope it helps.

Let me know if you have more questions.

Best,

Pablo

[ggoetten]

Hi @pablo-arantes

Thank you for the answers, I just noticed that the way the python script is handled, google colab starts on the correct stride after a crash or a disconnect. Very nice!

Thank you again!

Regards.