chore: migrate off v0 logic-env/math-env to v1 tasksets#2999
Draft
mikasenghaas wants to merge 2 commits into
Draft
chore: migrate off v0 logic-env/math-env to v1 tasksets#2999mikasenghaas wants to merge 2 commits into
mikasenghaas wants to merge 2 commits into
Conversation
These v0 environments are being removed from research-environments in favor of their v1 tasksets (logic-env -> i3_logic_v1, math-env -> math-env-v1, the latter already wired here). Remove the v0 editable path deps + relock so the build doesn't break when deps/research-environments is re-pinned past their deletion. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
Reviewed by Cursor Bugbot for commit 4cb47aa. Configure here.
| "ifeval-v1", | ||
| "lean-v1", | ||
| "livecodebench-v1", | ||
| "logic-env", |
There was a problem hiding this comment.
Configs still reference removed envs
Medium Severity
Dropping logic-env and math-env from the envs extra leaves many in-repo configs and examples still declaring id = "math-env" or id = "logic-env". After uv sync --extra envs, those runs fail to resolve the packages. math-env-v1 is already available as a successor, so the math configs are left pointing at a removed dep.
Additional Locations (1)
Reviewed by Cursor Bugbot for commit 4cb47aa. Configure here.
Move the training configs that used the v0 math-env on an explicit dataset to v1 env sections: gsm8k -> gsm8k-v1 (verifiers), and the Hendrycks/sanity configs -> math-env-v1 (task-facing math_verify_timeout preserved where set). All use the `null` harness + subprocess runtime. Drops the v0-only math_verify_max_workers. qwen30b_math and examples/Intellect-3.1 use the v0 math-env DEFAULT (INTELLECT-3-RL math), so they migrate to i3_math_v1 instead — deferred to the submodule bump. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
mikasenghaas
changed the title
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
Companion to PrimeIntellect-ai/research-environments#653 (the v0→v1 migration, which deletes the v0 envs). Moves prime-rl off the v0
logic-env/math-envso it keeps building oncedeps/research-environmentsis re-pinned past their deletion.logic-envandmath-enveditable path deps (from theenvslist and[tool.uv.sources]) + relock.math-env-v1stays; only the v0 envs go.taskset+nullharness +subprocessruntime; drops the v0-onlymath_verify_max_workers):configs/gsm8k/rl.toml→gsm8k-v1(from verifiers)configs/hendrycks_math,configs/nemotron_debug,configs/nemotron_4node,examples/multinode→math-env-v1(Hendrycks,math_verify_timeout=60preserved)configs/debug/hendrycks_sanity,examples/hendrycks_sanity→math-env-v1(sanity dataset)Why gsm8k/Hendrycks → math-env-v1 but not the others
math-env-v1is the Hendrycks-Math taskset; v0math-env's default is INTELLECT-3-RL math with difficulty filtering — a different env. Configs that passed an explicit Hendrycks/gsm8k dataset migrate cleanly; configs that relied on the INTELLECT-3-RL default do not map tomath-env-v1.Remaining before this can merge (staged on #653)
qwen30b_mathandexamples/Intellect-3.1use v0math-env's INTELLECT-3-RL default (and Intellect-3.1 also uses v0logic-env). Their correct v1 targets arei3_math_v1(being added in #653, parallel toi3_logic_v1) andi3_logic_v1— neither is in prime-rl's pinned submodule yet. So this PR stays draft until #653 merges, after which it will also:deps/research-environmentsto include chore(readme): fix fake command #653,i3-math-v1+i3-logic-v1path deps,qwen30b_math(→i3_math_v1) andexamples/Intellect-3.1(→i3_math_v1+i3_logic_v1), and relock.