docs: download the DAPO jsonl mirror the launcher expects - #2398
Merged
Conversation
Zhichenzzz
approved these changes
Aug 11, 2026
This was referenced Aug 11, 2026
yueming-yuan
added a commit
that referenced
this pull request
Aug 12, 2026
docs/getting-started/quick-start.md carried `<<<<<<< HEAD` / `>>>>>>> yueming/script-bugfix` from an earlier merge on this branch: that resolution pass checked for stale references but not for the markers themselves. The hunk kept the wrong side of #2398 as well, so the file now downloads zhuzilin/dapo-math-17k -- the mirror that actually carries the .jsonl the launcher reads -- into the --data-dir layout. `git grep '^<<<<<<< ' HEAD` over the whole tree finds nothing else.
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Fixes the Quick Start dataset download so the recipe actually runs end to end. A user following the page hit this: the DAPO download command fetched a repo whose contents don't match what
scripts/run-qwen3-4B.shconsumes.BytedTsinghua-SIA/DAPO-Math-17K, which today contains onlydata/dapo-math-17k.parquet, with the label nested atreward_model.ground_truth.--prompt-data /root/dapo-math-17k/dapo-math-17k.jsonlwith--input-key promptand a flat--label-key label, so the downloaded repo satisfied neither the path, the format, nor the key layout.zhuzilin/dapo-math-17kships exactly that file:dapo-math-17k.jsonlwhose rows are{"prompt": [chat messages], "label": "..."}(verified by fetching the repo tree and the first rows from the Hub). It is also the mirror every other recipe page already uses — including the AIME-2024 eval set on the same Quick Start page (zhuzilin/aime-2024), which is why AIME worked while DAPO didn't.Changes:
docs/getting-started/quick-start.md: downloadzhuzilin/dapo-math-17kinstead ofBytedTsinghua-SIA/DAPO-Math-17K.docs/models/qwen/index.md: point the DAPO-Math-17k link at the same mirror it names the file path of (link-only change).No launcher changes needed — with the right repo, the existing
--prompt-datapath and keys line up.The same audit found one more page with this class of bug (
docs/examples/reproducibility.mddownloadingopenai/gsm8k); that fix is in a follow-up PR.Test plan
hf download --repo-type dataset zhuzilin/dapo-math-17k --local-dir /root/dapo-math-17kproduces/root/dapo-math-17k/dapo-math-17k.jsonlprompt+ flatlabel, matching the launcher's keys