Make compute_average_normalized_field vacuum flag actually skip the plasma field calculation
#1576
Conversation
…acuum calculation for the compute_Bnormal call
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
dpanici
requested review from
a team,
rahulgaur104,
f0uriest,
ddudt,
sinaatalay,
unalmis and
YigitElma
and removed request for
a team
|
View / edit / reply to this conversation on ReviewNB YigitElma commented on 2025-02-08T17:25:36Z I assume this numerical difference comes from other changes and not this one? f0uriest commented on 2025-02-10T18:28:08Z This is likely the effect of analytically removing the plasma field (by only computing the vacuum field) vs keeping it, knowing it is numerically small |
|
View / edit / reply to this conversation on ReviewNB YigitElma commented on 2025-02-08T17:25:36Z These differences are due to changes we made to other parts of the code since the last notebook update, right? dpanici commented on 2025-02-11T15:31:50Z yea I did not realize it got this worse. I will redo the weights rn to see if can make it better |
|
View / edit / reply to this conversation on ReviewNB YigitElma commented on 2025-02-08T17:25:37Z This difference is easier to see |
|
View / edit / reply to this conversation on ReviewNB YigitElma commented on 2025-02-09T02:43:44Z Can you add these here too? I added them to each notebook except this one in #1539.
If you have access to a GPU, uncomment the following two lines before any DESC or JAX related imports. You should see about an order of magnitude speed improvement with only these two lines of code! # from desc import set_device
# set_device("gpu")
As mentioned in [DESC Documentation on performance tips](https://desc-docs.readthedocs.io/en/latest/performance_tips.html), one can use compilation cache directory to reduce the compilation overhead time. Note: One needs to create # import jax
# jax.config.update("jax_compilation_cache_dir", "../jax-caches")
# jax.config.update("jax_persistent_cache_min_entry_size_bytes", -1)
# jax.config.update("jax_persistent_cache_min_compile_time_secs", 0)
|
dpanici
changed the title
compute_average_normalized_field vacuum flag actually skip the vacuum calculationcompute_average_normalized_field vacuum flag actually skip the plasma field calculation
|
This is likely the effect of analytically removing the plasma field (by only computing the vacuum field) vs keeping it, knowing it is numerically small View entire conversation on ReviewNB |
|
yea I did not realize it got this worse. I will redo the weights rn to see if can make it better View entire conversation on ReviewNB |
…DESC into dp/coil-notebook-vac-flag
This saves about 10 seconds per call to the function for vacuum equilibria, which is like 20 seconds for the notebook