-
Notifications
You must be signed in to change notification settings - Fork 17
Ensemble Algebra Call 06 Dec 2016
Vivekanandan (Vivek) Balasubramanian edited this page Nov 27, 2018
·
2 revisions
On the way to O(1000) ensembles.
New set of data:
- 100k timesteps, 50 macrostates, 100 ensembles, dt=0.00002:
https://github.com/radical-collaboration/adaptive-msm/tree/master/entk/Data/iters1_100ksteps_50macros_100ensemble/plots
- 100k timesteps, 50 macrostates, 100 ensembles, dt=0.0001:
https://github.com/radical-collaboration/adaptive-msm/tree/master/entk/Data/iters_1_100k_timesteps_50_macros/plots
New set of data:
All data is available on Stampede at /work/02734/vivek91/expanded_ensemble_data
.
- "fixed_80k_3iters_attempt_1" - 80K timesteps per simulation, 3 iterations
- "fixed_80k_5iters_attempt_1" - 80K timesteps per simulation, 5 iterations
- "fixed_80k_10iters_attempt_1" - 80K timesteps per simulation, 10 iterations
- "fixed_100k_5iters_attempt_1" - 100K timesteps per simulation, 5 iterations
- "fixed_100k_10iters_attempt_1" - 100K timesteps per simulation, 10 iterations
Currently, the data of interest (kept at the above location) is about ~350GB. Requested additional space (500-1000GB) on Jetstream to house this data.
Earlier error recreated, files logged and stored.
- Failed task:
/work/02734/vivek91/radical.pilot.sandbox/rp.session.js-168-20.jetstream-cloud.org.vivek91.017139.0001-pilot.0000/unit.000471
- Data from previous iterations:
/work/02734/vivek91/expanded_ensemble_data/failed_attempt
- Added pseudographs to the paper. Specifically: graph investigating parameter and resource adaptivity, strong and weak scalability graphs, convergence as a function of walltime
- Tabularized adaptivity
- Debug error in EE.
- Keep going bigger with MSM experiments.
- Investigate how to design and implement a) resource adaptivity experiments and b) convergence vs walltime experiments
- Reiterate on adaptivity and experiment section, rest might be more straightforward to write.
- Will be debugging error in EE usecase with Michael
- Look into the MSM papers to check whether the results we got are expected