Skip to content

chore(beep boop 🤖): Bump uv.lock (r0.5.0, mcore-core_r0.18.0) (2026-07-02)#4627

Open
svcnvidia-nemo-ci wants to merge 1 commit into
r0.5.0from
bump-ci-container-2026-07-02-r0.5.0-core_r0.18.0
Open

chore(beep boop 🤖): Bump uv.lock (r0.5.0, mcore-core_r0.18.0) (2026-07-02)#4627
svcnvidia-nemo-ci wants to merge 1 commit into
r0.5.0from
bump-ci-container-2026-07-02-r0.5.0-core_r0.18.0

Conversation

@svcnvidia-nemo-ci

Copy link
Copy Markdown
Contributor

🚀 PR to bump uv.lock in r0.5.0.

🤖 This PR will be merged automatically once CI passes.

…-07-02)

Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
@svcnvidia-nemo-ci

Copy link
Copy Markdown
Contributor Author

/ok to test c179c3e

@copy-pr-bot

copy-pr-bot Bot commented Jul 2, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@yaoyu-33

yaoyu-33 commented Jul 2, 2026

Copy link
Copy Markdown
Contributor

MCore bump auto-fix status for release-r0.5.0:

Classification: Bridge broke itself
Evidence: On 2026-07-02, CI run https://github.com/NVIDIA-NeMo/Megatron-Bridge/actions/runs/28584202436 failed only L2_Launch_models_qwen_quantization (job https://github.com/NVIDIA-NeMo/Megatron-Bridge/actions/runs/28584202436/job/84772073239) and gb200_L2_Launch_models_qwen_quantization (job https://github.com/NVIDIA-NeMo/Megatron-Bridge/actions/runs/28584202436/job/84772073154). Both logs fail in modelopt/torch/quantization/plugins/transformer_engine.py:178 with TypeError: object of type 'bool' has no len() after MCore calls grouped linear with m_splits. The r0.5.0 base already combines TransformerEngine b9d690e042b1c4e455214e7dab65d6d3512c05d6 with nvidia-modelopt==0.44.0rc5 through merged PR #4535 from 2026-06-26. The MCore range d30c93ffae858b22eece3fa71c734c8f43161eff...458c8d0ecafdf6d9e36771600d62ade27f2a67b7 is two commits and changes only MCore's TransformerEngine dependency metadata plus uv.lock.
Fix PR: not opened. The directly relevant release fix #4615 used the same target MCore SHA, restored the ModelOpt-compatible TransformerEngine revision, and was closed unmerged on 2026-07-01. PR #4600 is open against main, has green H100 and GB200 Qwen quantization checks, but has no r0.5.0 backport label and does not itself update the release branch.
Guards: none added or removed; this is a dependency-pin compatibility issue, not a code-guard case.
Validation: PR #4627's import, unit, lint, installation, and all functional checks except the two Qwen3 MoE quantization jobs passed on 2026-07-02. No new CW interactive validation was run because no replacement fix is authorized while #4615 remains closed unmerged. Prior #4615 validation on 2026-07-01 passed uv lock --check, 81 focused unit tests, and a grouped-linear compatibility smoke test in CW interactive job 13313466.
Next action: maintainer decision needed — either authorize reopening/rebasing #4615 onto #4627, or merge #4600 and explicitly backport its dependency rollback to r0.5.0. Until that decision, do not merge #4627 because both H100 and GB200 Qwen3 MoE quantization remain red.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants