---
license: mit
---

# SWE-Bench Trajectory Eval Bundle (v2)

Airgapped-ready bundle for the trajectory-probe downstream eval.

## Changes from v1

- **django dropped** (230 base_commits × 5 agents = 1139 trajectories removed).
  django's blobs were too slow to prefetch at scale; the other 11 repos
  give 1317 trajectories which is plenty.
- **Sparse-checkout blobs only**. For each (repo, base_commit), only the
  files any trajectory's patch touches are fetched. Dramatically smaller
  than the v1 "full-tree" approach; still enough for the graph encoder
  to see the patched files + their containing package.

## Contents (570 MB compressed)

- `specs.jsonl`: 1317 SWE-Bench Verified agent trajectories (down from
  2456). 46% resolved. 11 repos, 269 unique base_commits.
- `repos/`: 11 blobless clones with sparse-checkout already configured
  per base_commit. Working trees on-disk reflect the UNION of touched
  paths across all trajectories for each base_commit.
- `graphjepa/`: pipeline code (trajectory_pipeline with `extract_patch_paths`
  and `set_sparse_paths`, trajectory_realize, trajectory_probe, checks).
- `harvest.log`.

## Dataset

| Repo | Trajs | Resolved |
|---|---|---|
| sympy/sympy | 370 | 48% |
| sphinx-doc/sphinx | 207 | 32% |
| matplotlib | 167 | 41% |
| scikit-learn | 158 | 70% |
| astropy | 110 | 42% |
| pydata/xarray | 108 | 51% |
| pytest-dev/pytest | 95 | 49% |
| pylint-dev/pylint | 49 | 22% |
| psf/requests | 38 | 61% |
| mwaskom/seaborn | 10 | 0% |
| pallets/flask | 5 | 80% |
| **total** | **1317** | **46%** |

## Downstream workflow (on airgapped big)

```bash
huggingface-cli download IDMedicine/code-graph-trajeval-v2 traj_full_bundle.tar.gz \
    --local-dir outputs/traj_real

cd /opt/train/code-transformer/code-transformer
tar -xzf outputs/traj_real/traj_full_bundle.tar.gz
rsync -a traj_full/graphjepa/ graphjepa/
mkdir -p outputs/traj_real
cp traj_full/specs.jsonl outputs/traj_real/
mv traj_full/repos outputs/traj_real/repos

export BERT_MODEL_PATH=/opt/train/datasets/bert-base-uncased
GPUS=0,2,3,4,5,6,7 bash graphjepa/scripts/trajectory_full.sh
tail -f outputs/traj_real/logs/prebuild_shard*_gpu*.log
```

Because the clone already has the right blobs prefetched on laptop,
the airgapped big machine can checkout every base_commit without any
network access.