Back To Multistage Rosetta Scripts
One of the great things about traditional rosetta scripts is that there is very little penalty for crashing. If the program crashes 80% of the way through, you still get 80% of the results and you can just re-submit the job to get the final 20%.
Multistage Rosetta Scripts is not like that! If the program dies for any reason before the structures are output, you get NOTHING!
To counter this, we plan on implementing a checkpointing system that allows progress to be saved and loaded in the event of job failure.
Everything below this point is pure speculation.
<PROTOCOLS>
<Stage total_num_results_to_keep="1000">
<Add mover="my_fast_design_mover"/>
<Sort filter="my_sfxn_filter"/>
</Stage>
<Checkpoint filename="cp1"/>
<Stage total_num_results_to_keep="50">
<Add mover="any_mover"/>
<Sort filter="any_filter"/>
</Stage>
<Stage total_num_results_to_keep="30">
<Add mover="any_mover"/>
<Sort filter="any_filter"/>
</Stage>
<Checkpoint filename="cp2" old_filename_to_delete="cp1"/>
<Stage total_num_results_to_keep="10">
<Add mover="any_mover"/>
<Sort filter="any_filter"/>
</Stage>
<PROTOCOLS>TODO write about this
<PROTOCOLS>
- <Stage total_num_results_to_keep="1000">
+ <Add mover="my_fast_design_mover"/>
+ <Sort filter="my_sfxn_filter"/>
</Stage>
- <Cluster total_num_results_to_keep="60">
+ <Sequence/>
- </Cluster>
- <Checkpoint filename="cp1"/>
- <Stage total_num_results_to_keep="50">
+ <Add mover="any_mover"/>
+ <Sort filter="any_filter"/>
</Stage>
- <Stage total_num_results_to_keep="30">
+ <Add mover="any_mover"/>
+ <Sort filter="any_filter"/>
</Stage>
- <Checkpoint filename="cp2" old_filename_to_delete="cp1"/>
- <Stage total_num_results_to_keep="10">
+ <Add mover="any_mover"/>
+ <Sort filter="any_filter"/>
- </Stage>
<PROTOCOLS>