Repository navigation
SWE training sandbox migration - #3327
Merged
Merged
Conversation
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
yaoyu-33
added
area:environment
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
…box-migration Signed-off-by: Sugam Devare <sdevare@nvidia.com>
Signed-off-by: Sugam Devare <sdevare@nvidia.com>
…ripts Signed-off-by: Sugam Devare <sdevare@nvidia.com>
These two files were copied in incidentally from denovo_swe/.gitignore in the "reward profiling" commit. With no pyproject.toml/requirements.txt, CI's server-test matrix picks up resources_servers/openswe as a changed server and fails trying to build its venv. openswe has no actual server implementation on this branch; that work is stashed separately. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Sugam Devare <sdevare@nvidia.com>
…_pro rows) - openswe row: dead link, follow-up to the stub-file removal in this branch. - swebench_pro row: pre-existing drift, unrelated to this branch -- verified it reproduces identically on a clean origin/main checkout. That server's config has no agent-side `datasets` block, so update_env_list.py's types-based inclusion check has always excluded it; README just hadn't been regenerated since. Table generation via update-readme-table always rewrites the row for both cases together, so committing them separately isn't possible without --no-verify. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Sugam Devare <sdevare@nvidia.com>
…box-migration Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Signed-off-by: Hemil Desai <hemildesai@users.noreply.github.com>
gym env test only tests modules with a README.md, so the server suite's total-vs-tested module check failed (195 found, 193 tested) for these two new servers. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Signed-off-by: Hemil Desai <hemildesai@users.noreply.github.com>
Contributor
|
/ok to test e8fbc85 |
Contributor
|
/ok to test c4ca745 |
bxyu-nvidia
approved these changes
Sep 23, 2026
This branch was previously deployed
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.
SWE training sandbox migration
Summary
Adds/updates four SWE-bench-style datasets for RL training, each graded via a golden-patch 3x
full-set sweep (grade every row with its own patch, keep only what resolves in every pass):
20,181 Python tasks. 17,696 resolve in all 3 passes.
32,079 tasks across 20 languages. 22,663 resolve in all 3 passes.
swe_bench_extinstead of a hand-rolled parser. 8,726 of 9,601 raw candidates resolve in all 3passes.
swe_bench_extmechanism as swemer_v1, confirmed to grade identically (149/149 sample) beforereplacing the old parsers.
data/supported_instance_ids.txtrebuilt against it: 2,515 of 2,538resolve in all 3 passes.