Correct the resource labels from the run trace, and speed up lm - #58
Merged
Conversation
Every column of mod2 is regressed on the same design matrix, but the script called `fastLm()` once per column inside a serial `pblapply()`, redoing an n x 51 decomposition for each of the ~229k ATAC peaks. Solving the normal equations once gives the identical fit -- on synthetic data of that shape the predictions agree to 2e-15, also when the design matrix is badly conditioned. In run 1gVHaBsNbAHcfT lm was the only method near its walltime, at 7.6h of 8h. * Replace the per-column loop with a single `solve()` * Drop the now unused RcppArmadillo and pbapply dependencies
de.NBI Berlin has two GPU flavours, both a single T4: `de.NBI GPU T4 medium` (16 vCPU / 64 GB) and `de.NBI GPU T4 large` (32 vCPU / 128 GB). In run 1gVHaBsNbAHcfT these four steps peaked at 1.4-3.5 GB while each occupying a whole large node, and they were 20 of the ~35 jobs queued behind the two T4s. * novel_predict, simple_mlp_predict, ss_opm_predict, babel_predict: gpu -> midgpu The train steps stay on `gpu`: ss_opm_train really does peak at 96 GB.
* knnr_py, cellmapper_linear, knnr_r: midmem -> highmem * cellmapper_linear, knnr_r: midcpu -> lowcpu * solution, zeros: midmem -> lowmem knnr_py peaked at 82 GB and cellmapper_linear at 68 GB against their 50 GB request, and both were OOM-killed once in run 1gVHaBsNbAHcfT; knnr_r sits at 45 GB of 50. The other way round, solution uses 0.5 GB and zeros 11 GB of their 50 GB, and knnr_r and cellmapper_linear use 1.4 and 3.8 of 15 cores.
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Describe your changes
Corrects the resource labels against the trace of run
1gVHaBsNbAHcfT, and rewriteslm's inner loop rather than giving it more walltime.Checklist before requesting a review
I have performed a self-review of my code
Check the correct box. Does this PR contain:
Proposed changes are described in the CHANGELOG.md
CI Tests succeed and look good!