feat: schedule resident inference batches - #53
Draft
DimaMolod wants to merge 9 commits into
Draft
Conversation
This was referenced Sep 3, 2026
…sident AlphaFold2 compiles per input shape, so a resident batch whose folds differ in length recompiles for each one and saves nothing. A resident batch also keeps its compiled model in memory, so the on-disk compile cache it was being given is redundant there. - add --desired_num_res for AF2 multimer batches, sized from the batch's largest fold, so the batch compiles once - build a separate resident CLI without --jax_compilation_cache_dir - accept --jax_compilation_cache_dir on AF2 in the parse-time flag check, so it can be set explicitly in structure_inference_arguments Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
config.yaml had grown to 248 lines of which 180 were comments, making the ~34 actual settings hard to find. - group config.yaml into REQUIRED / COMMON / ADVANCED, one or two comment lines per key naming the README section that documents it (248 -> 134 lines, every key and default unchanged) - document the three sections in README - describe AF2 shape padding in the batching section, and drop the calibration caveat about resident allocator retention - correct the claim that --jax_compilation_cache_dir is AlphaFold3-only, and note the network-filesystem failure mode Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
batch_padding_args was reading staged features, the same instability that _fold_grouping_tokens exists to avoid, and an unresolved length could have padded a batch to fewer residues than its largest fold. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The copied table omitted convert_to_modelcif, so setting a valid flag on AlphaFold3 produced a "backend does not recognise" warning. AlphaPulldown is not importable at parse time, so the copy has to stay; record where truth lives and pin the sets in a test, since this has drifted once already. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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.
Summary
bin_foldsgroupingTests
Depends on and complements KosinskiLab/AlphaPulldown#636.
Related MMseqs2-GPU workflow PR: #54.