Cassandra 1.2 > Restarting Simulations
- piskuliche
-
Topic Author
- Offline
- Junior Member
-
Less
More
- Posts: 30
- Thank you received: 1
7 years 10 months ago #132
by piskuliche
Hello,
I have a quick question. I have noticed that version 1.2 did not include the read_old.py script with it, is this because we should start moving over to starting from the checkpoint files?
With that in mind does the checkpoint file override
> box size
> max displacements for moves
etc that the old read_old.py script used to do?
Thanks!
I have a quick question. I have noticed that version 1.2 did not include the read_old.py script with it, is this because we should start moving over to starting from the checkpoint files?
With that in mind does the checkpoint file override
> box size
> max displacements for moves
etc that the old read_old.py script used to do?
Thanks!
Please Log in to join the conversation.
- ryangmullen
-
- Offline
- Administrator
-
Less
More
- Posts: 124
- Karma: 4
- Thank you received: 24
7 years 10 months ago #133
by ryangmullen
Hi Zeke,
Restarting from a checkpoint file uses the configuration, box size, maximum displacement for moves and random seeds from the end of the previous run. The read_old.py script duplicated all of that functionality so we did not update it for Cassandra 1.2.
Cheers,
Ryan
Restarting from a checkpoint file uses the configuration, box size, maximum displacement for moves and random seeds from the end of the previous run. The read_old.py script duplicated all of that functionality so we did not update it for Cassandra 1.2.
Cheers,
Ryan
Please Log in to join the conversation.
- piskuliche
-
Topic Author
- Offline
- Junior Member
-
Less
More
- Posts: 30
- Thank you received: 1
7 years 10 months ago #134
by piskuliche
Hi Ryan,
My coworkers and I have noticed an interesting problem with 1.2 with the binary GEMC.
When I try to restart one of my simulations (run with 1.1 of Cassandra) from the checkpoint file, the simulation starts and then immediately moves to "Cassandra Simulation Complete" without respect to how many steps/sweeps that I choose. Even with logfile verbose it doesn't give any explanation of what is happening.
I have included the input files that produce this problem.
Any idea what might be going on?
Zeke
My coworkers and I have noticed an interesting problem with 1.2 with the binary GEMC.
When I try to restart one of my simulations (run with 1.1 of Cassandra) from the checkpoint file, the simulation starts and then immediately moves to "Cassandra Simulation Complete" without respect to how many steps/sweeps that I choose. Even with logfile verbose it doesn't give any explanation of what is happening.
I have included the input files that produce this problem.
Any idea what might be going on?
Zeke
Please Log in to join the conversation.
- ryangmullen
-
- Offline
- Administrator
-
Less
More
- Posts: 124
- Karma: 4
- Thank you received: 24
7 years 10 months ago - 7 years 10 months ago #135
by ryangmullen
Hi Zeke,
The checkpoint file lists the final MC step of the simulation that generated the checkpoint file. This becomes the initial MC step of any simulations started from the checkpoint, initial_mcstep.
In the input file, you list the final MC step to run to with the keyword 'run', n_mcsteps.
If initial_mcstep >= n_mcsteps then Cassandra computes the energy of the input configuration and quits.
For example, if you run a simulation for 100 steps, the final checkpoint file will list the MC step as 100. If try to restart a simulation from that checkpoint using the 'run 100' command in your input file, Cassandra sees a request to run 100 steps, but also sees that you are already on step 100.
______________________________________________
Maybe we should have separate 'run' and 'run_to' keywords. What do you think?
Ryan
The checkpoint file lists the final MC step of the simulation that generated the checkpoint file. This becomes the initial MC step of any simulations started from the checkpoint, initial_mcstep.
In the input file, you list the final MC step to run to with the keyword 'run', n_mcsteps.
If initial_mcstep >= n_mcsteps then Cassandra computes the energy of the input configuration and quits.
For example, if you run a simulation for 100 steps, the final checkpoint file will list the MC step as 100. If try to restart a simulation from that checkpoint using the 'run 100' command in your input file, Cassandra sees a request to run 100 steps, but also sees that you are already on step 100.
______________________________________________
Maybe we should have separate 'run' and 'run_to' keywords. What do you think?
Ryan
Last Edit: 7 years 10 months ago by ryangmullen.
Please Log in to join the conversation.
- piskuliche
-
Topic Author
- Offline
- Junior Member
-
Less
More
- Posts: 30
- Thank you received: 1
7 years 10 months ago #136
by piskuliche
Thanks for the reply.
That would work, however I think it also is indicative of a weird interaction between having previously run stuff in units steps and now switching over to sweeps. (i.e. when I restart a simulation with 1000 molecules that had only done 430 sweeps (430,000 steps) and tell it to run 2000 total sweeps it sees the 2000 is less than 430,000 and doesn't do the sweeps (when really it should be comparing 2000 * steps_per_sweep with the 430,000 value.)
I'm not completely positive this is what is happening, but that is how it seems to me.
That would work, however I think it also is indicative of a weird interaction between having previously run stuff in units steps and now switching over to sweeps. (i.e. when I restart a simulation with 1000 molecules that had only done 430 sweeps (430,000 steps) and tell it to run 2000 total sweeps it sees the 2000 is less than 430,000 and doesn't do the sweeps (when really it should be comparing 2000 * steps_per_sweep with the 430,000 value.)
I'm not completely positive this is what is happening, but that is how it seems to me.
Please Log in to join the conversation.
- ryangmullen
-
- Offline
- Administrator
-
Less
More
- Posts: 124
- Karma: 4
- Thank you received: 24
7 years 10 months ago #137
by ryangmullen
Looking at your files, the checkpoint file lists the MC step as 200,000,000
The input file says to run 200,000 sweeps, but steps_per_sweep is undefined, i.e. keyword 'steps_per_sweep' has no option, e.g. 'steps_per_sweep 1000', so it defaults to 'steps_per_sweep 1'.
So Cassandra sees you are on step 200,000,000 and that you said to run to 200,000 sweeps * 1 step/sweep = 200,000 steps.
The input file says to run 200,000 sweeps, but steps_per_sweep is undefined, i.e. keyword 'steps_per_sweep' has no option, e.g. 'steps_per_sweep 1000', so it defaults to 'steps_per_sweep 1'.
So Cassandra sees you are on step 200,000,000 and that you said to run to 200,000 sweeps * 1 step/sweep = 200,000 steps.
Please Log in to join the conversation.
Time to create page: 0.129 seconds