Re-enable test_qwen3_4B_ckpt.py - #1271
Merged
Merged
Conversation
This checkpoint save/load hash-equality test was disabled with disabled="Disabled due to bugs." It passes today: verified twice on an 8xH200 devbox (image radixark/miles:dev) via the exact CI invocation `python3 tests/e2e/ckpt/test_qwen3_4B_ckpt.py`, both runs deterministic (Model hashes match x4, peak GPU 12.37 GB, zero assertion failures). Removes the FIXME and the disabled marker; no assert logic is touched.
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request re-enables the Qwen3 4B checkpoint end-to-end test by removing the disabled parameter and the associated FIXME comment from the register_cuda_ci registration. There are no review comments, so I have no feedback to provide.
yueming-yuan
approved these changes
Jun 3, 2026
guapisolo
added a commit
that referenced
this pull request
Jul 12, 2026
Disabled in #1149 as "Disabled due to bugs" with a FIXME pointing at Megatron. Since then the Megatron pin picked up the dist-ckpt load fix for bare BytesIO _extra_state (radixark/Megatron-LM#53, fixes #1293), and the sibling qwen3_4B ckpt test disabled in the same sweep proved stale on re-run (#1271). The training config itself (TP2/PP2/CP2/EP4 + MTP + EAGLE) already runs green in test_glm47_flash/test_r3_mtp; the save -> load -> async_save -> load roundtrip is the only untested part. Re-enable and let the run-ci-ckpt CI run on 8xH100 validate it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
guapisolo
added a commit
that referenced
this pull request
Jul 14, 2026
Disabled in #1149 as "Disabled due to bugs" with a FIXME pointing at Megatron. Since then the Megatron pin picked up the dist-ckpt load fix for bare BytesIO _extra_state (radixark/Megatron-LM#53, fixes #1293), and the sibling qwen3_4B ckpt test disabled in the same sweep proved stale on re-run (#1271). The training config itself (TP2/PP2/CP2/EP4 + MTP + EAGLE) already runs green in test_glm47_flash/test_r3_mtp; the save -> load -> async_save -> load roundtrip is the only untested part. Re-enable and let the run-ci-ckpt CI run on 8xH100 validate it. Co-Authored-By: Claude Fable 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
3 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
Re-enables
tests/e2e/ckpt/test_qwen3_4B_ckpt.py, which was disabled withdisabled="Disabled due to bugs."and a# FIXME: fix thiscomment.The test exercises Megatron checkpoint save → load round-trips (sync and async save) for Qwen3-4B under TP2/PP2/CP2 + sequence-parallel, and asserts per-
(tp,pp,dp,cp)-rank SHA256 hash-equality of model parameters across the boundary (--ci-save-model-hashon save,--ci-check-model-hash+--low-memory-resumeon load). The hash check runs immediately after load, before the load-run's training loop, so the assertion is deterministic w.r.t. rollout non-determinism. No assert logic is touched — only the registration marker is restored.Test plan
radixark/miles:dev) via the exact CI invocationpython3 tests/e2e/ckpt/test_qwen3_4B_ckpt.py(no args).Model hashes match×4 (two save/load round-trips: sync + async),Saved model hashes×8 (8 ranks), all 4 Ray jobssucceeded, zeroAssertionError/ hash mismatch / traceback.--low-memory-resumepeak-GPU-memory guard satisfied: 12.37 GB < 20 GB threshold on every rank.