Repository navigation
Release Cerberus models - #1
Merged
Merged
Conversation
Add releases/cerberus: download script, params, targets, and verification fixtures for 8 folds of the joint human/mouse Cerberus model, hosted at gs://seqnn-share/cerberus/. Generalize hound_verify to joint-model families (one multi-species model per fold, head selected by species) and run Hydra models on CUDA, since the Hydra scan is a Triton kernel. Refocus the README on Cerberus. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Wiz Scan Summary
To detect these findings earlier in the dev lifecycle, try the Wiz Code extension for VS Code, JetBrains, or Visual Studio. |
Co-Authored-By: Claude Opus 5.5 <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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Releases the 8 Cerberus replicate models. Each one is a single model with a human head (8,361 tracks) and a mouse head (3,102 tracks), taking 786 kb input at 32 bp resolution. The weights are hosted at
gs://seqnn-share/cerberus/.releases/cerberus/:download.sh(./download.sh [FOLD|all] [DEST])params.json, which includes thetrainrecipetargets_{human,mouse}.txtverify/fixtures: synthetic weights, plus each fold × species headhound_verifynow supports joint families. A family is joint whenreleases/<family>/params.jsonexists, and its weights live atmodels/f<n>c0/instead ofmodels_<species>/.test_published_forwardskips CUDA-only families on hosts without a GPU.The training data release is pending, and the Cerberus README says so.
Test plan
hound_verify --family cerberuson GPU with all 8 staged folds: 16/16 PASS (r=1.0, rmse_rel=0). Repeated runs give identical results.hound_verify --family cerberuswith no weights downloaded (synthetic): PASS for both heads../download.sh 0, thenhound_verify --family cerberus, on a GPU node: PASS for both heads.hound_verify --family borzoi/borzoi_primeon CPU (synthetic): PASS, so the existing fixtures are unchanged.pytest src/tests/test_verify.py: 6 passed, 2 skipped (Cerberus on a CPU host).🤖 Generated with Claude Code