chore(beep boop 🤖): Bump uv.lock (main, mcore-main) (2026-04-08) - #3210
Conversation
Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
|
/ok to test 42ad86b |
📝 WalkthroughWalkthroughThis PR updates the Megatron-LM submodule reference to a new commit hash. The Changes
Estimated code review effort🎯 1 (Trivial) | ⏱️ ~2 minutes Possibly related PRs
Suggested reviewers
🚥 Pre-merge checks | ✅ 3 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.main.commit:
- Line 1: The PR title incorrectly says "Bump `uv.lock`" while the change
actually updates the Megatron-LM submodule reference in .main.commit (commit
8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8); update the PR title and commit
message to accurately describe the change (e.g., "Bump Megatron-LM submodule to
8cf6b35") or, if the intent was to bump uv.lock, revert this submodule change
and make the uv.lock update instead so PR title and .main.commit stay
consistent.
- Line 1: The .main.commit entry points to a non-existent Megatron-LM commit
(8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8), which will break the
3rdparty/Megatron-LM submodule; verify whether that commit was pushed to the
Megatron-LM repo or the hash is mistyped, then update the submodule pointer and
.main.commit to a valid commit: either push the missing commit to the
Megatron-LM remote or replace the hash in .main.commit and run git submodule
update --init --recursive (and commit the corrected submodule reference) so
3rdparty/Megatron-LM points to an existing commit.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: a74486f5-0b92-4ae1-a993-360b78f1a5e3
⛔ Files ignored due to path filters (1)
uv.lockis excluded by!**/*.lock
📒 Files selected for processing (2)
.main.commit3rdparty/Megatron-LM
| @@ -1 +1 @@ | |||
| 17a67b9a97fb11a75933fd7f76ad76e1ac98a53d No newline at end of file | |||
| 8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8 | |||
There was a problem hiding this comment.
Note: PR title mentions uv.lock but changes are to submodule reference.
The PR title states "Bump uv.lock" but the actual changes update the Megatron-LM submodule reference in .main.commit. This may cause confusion when reviewing the PR history.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In @.main.commit at line 1, The PR title incorrectly says "Bump `uv.lock`" while
the change actually updates the Megatron-LM submodule reference in .main.commit
(commit 8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8); update the PR title and
commit message to accurately describe the change (e.g., "Bump Megatron-LM
submodule to 8cf6b35") or, if the intent was to bump uv.lock, revert this
submodule change and make the uv.lock update instead so PR title and
.main.commit stay consistent.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Description: Verify the commit exists in Megatron-LM and check submodule consistency
# Check if the commit exists in the submodule
cd 3rdparty/Megatron-LM
if git cat-file -e 8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8 2>/dev/null; then
echo "✓ Commit exists in Megatron-LM repository"
git log -1 --oneline 8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8
else
echo "✗ Commit not found in Megatron-LM repository"
fi
cd ../..
# Verify submodule pointer matches .main.commit
SUBMODULE_HASH=$(git ls-tree HEAD 3rdparty/Megatron-LM | awk '{print $3}')
MAIN_COMMIT_HASH=$(cat .main.commit | tr -d '[:space:]')
echo ""
echo "Submodule pointer: $SUBMODULE_HASH"
echo ".main.commit hash: $MAIN_COMMIT_HASH"
if [ "$SUBMODULE_HASH" = "$MAIN_COMMIT_HASH" ]; then
echo "✓ Submodule pointer matches .main.commit"
else
echo "✗ Mismatch between submodule pointer and .main.commit"
fiRepository: NVIDIA-NeMo/Megatron-Bridge
Length of output: 278
Commit hash does not exist in the Megatron-LM repository—this will break the submodule.
The commit 8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8 is not found in the 3rdparty/Megatron-LM submodule, despite the submodule pointer and .main.commit file both referencing it consistently. This means users cloning this repository will fail to fetch the submodule.
Verify that:
- The commit has been pushed to the Megatron-LM repository, or
- The commit hash is correct
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In @.main.commit at line 1, The .main.commit entry points to a non-existent
Megatron-LM commit (8cf6b355a7bd8fe6abe39bc7596bd3c1cb92bee8), which will break
the 3rdparty/Megatron-LM submodule; verify whether that commit was pushed to the
Megatron-LM repo or the hash is mistyped, then update the submodule pointer and
.main.commit to a valid commit: either push the missing commit to the
Megatron-LM remote or replace the hash in .main.commit and run git submodule
update --init --recursive (and commit the corrected submodule reference) so
3rdparty/Megatron-LM points to an existing commit.
…heckpoint layout Real GPT-OSS checkpoints store down_proj as [num_experts, hidden, intermediate] but HF model init produces [num_experts, intermediate, hidden]. After PR #3162 removed the import-side transpose, the CI toy models need to match the real layout. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
|
/ok to test af7d51b |
GPT-OSS down_proj is intentionally transposed on export (megatron_to_hf) per PR #3162. Align shapes before torch.allclose to avoid RuntimeError when the exported param has swapped dimensions vs the original. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
52c6e7f to
af009ac
Compare
|
/ok to test af009ac |
🚀 PR to bump
uv.lockinmain.🤖 This PR will be merged automatically once CI passes.
Summary by CodeRabbit
Note: This release contains no user-facing feature changes or bug fixes. The updates are internal maintenance-related adjustments.