Precompute symbolic matrix to hold jacobian #41
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jumpinthefire
force-pushed
the
faer-exp
branch
from
af20b1e to
63d912c
Compare
powei-lin
reviewed
Jan 20, 2025
src/problem.rs
Outdated
| for (i, var_key) in residual_block.variable_key_list.iter().enumerate() { | ||
| if let Some(variable_global_idx) = variable_name_to_col_idx_dict.get(var_key) { | ||
| if let Some(_) = variable_name_to_col_idx_dict.get(var_key) { |
There was a problem hiding this comment.
If this is not being used anymore, can you change it to if contains_key else panic (because it should be here) ?
powei-lin
reviewed
Jan 20, 2025
powei-lin
reviewed
Jan 20, 2025
|
There is a printlin in LM. My mistake. Can you help me change that to trace, please? |
|
I tried and it's like ~10% faster (depends on how large and how empty is the jacobian). Amazing! |
Done ✔️ |
# for free
to join this conversation on GitHub.
Already have an account?
# to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR allows the optimizer to pre-compute a symbolic matrix to hold the Jacobian. This ends up consistently shedding time at each iteration of the optimization loop, across all examples, by invoking
SparseColMat::new_from_order_valuesinstead ofSparseColMat::try_new_from_triplets.Since the original insertion order must be respected to make use of the symbolic matrix, we've moved away from a shared list of jacobian values (which may have items inserted into it in any order) to each residual block returning a sublist of values. These sublists are then flattened in the same order as the residual blocks.
Note: compiling the tests in CI/CD was failing because of the plotting library, installing the right native dependencies seems to fix it.