chore: bump mbridge and MLM - #2354
Conversation
Signed-off-by: ashors1 <ashors@nvidia.com>
Signed-off-by: Anna Shors <ashors@nvidia.com>
✅ Submodule Fast-Forward Check ResultsCheck based on commit: c4e187e (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
✅ Submodule Fast-Forward Check ResultsCheck based on commit: ba19dfb (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
Signed-off-by: ashors1 <ashors@nvidia.com>
✅ Submodule Fast-Forward Check ResultsCheck based on commit: 9e8469e (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
✅ Submodule Fast-Forward Check ResultsCheck based on commit: 1e4f42a (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
Signed-off-by: ashors1 <ashors@nvidia.com>
✅ Submodule Fast-Forward Check ResultsCheck based on commit: 0f3a50f (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
Signed-off-by: ashors1 <ashors@nvidia.com>
✅ Submodule Fast-Forward Check ResultsCheck based on commit: f0535cc (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
Signed-off-by: ashors1 <ashors@nvidia.com>
✅ Submodule Fast-Forward Check ResultsCheck based on commit: 3061928 (PR #2354 from ✅ Submodules that are properly updated:Megatron-Bridge: ✅ PR branch is ahead of main branch (fast-forward) All submodule changes look good! ✨ |
|
@ashors1 Thanks for the great work! Just wondering — do you have a rough idea of when the PR might be merged? |
Signed-off-by: ashors1 <ashors@nvidia.com>
Signed-off-by: ashors1 <ashors@nvidia.com>
@yuekaizhang Sorry for the late reply. I am currently running tests to make sure the bump doesn't introduce any regressions. If the tests pass, hopefully we can get the PR in very soon. But if the tests fail, I might need to spend a few days debugging. We have many PRs we are trying to get in a the moment and this one is not the highest priority unfortunately. |
Signed-off-by: Anna Shors <ashors@nvidia.com>
Signed-off-by: Anna Shors <ashors@nvidia.com>
Bump the Megatron-Bridge submodule to 7918b0f2 (nested Megatron-LM to 38986a98), mirroring the dependency changes from PR NVIDIA-NeMo#2354 (ashors/bump-mbridge-mlm): - transformers 5.3 -> 5.8.1,<5.9.0 in the megatron-bridge proxy CACHED_DEPENDENCIES and root static dependency-metadata - drop transformer-engine / mamba-ssm / causal-conv1d from the megatron-bridge dep set (now extras upstream); move mamba-ssm + causal-conv1d to the mcore extra - add mistral-common, flashinfer-python, flashinfer-cubin - bump nvidia-resiliency-ext rev and logsage metadata to 0.1.7 - regenerate uv.lock (PR lock minus pytest-shard, which our main drops) - Dockerfile: make z3-solver libz3 discoverable for fla/tilelang - modelopt worker: register "nemo_rl." as an mbridge allowed target prefix (new API in this Megatron-Bridge) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Zhaopeng Qiu <alexq@nvidia.com>
Bump the Megatron-Bridge submodule to 7918b0f2 (nested Megatron-LM to 38986a98), mirroring the dependency changes from PR NVIDIA-NeMo#2354 (ashors/bump-mbridge-mlm): - transformers 5.3 -> 5.8.1,<5.9.0 in the megatron-bridge proxy CACHED_DEPENDENCIES and root static dependency-metadata - drop transformer-engine / mamba-ssm / causal-conv1d from the megatron-bridge dep set (now extras upstream); move mamba-ssm + causal-conv1d to the mcore extra - add mistral-common, flashinfer-python, flashinfer-cubin - bump nvidia-resiliency-ext rev and logsage metadata to 0.1.7 - regenerate uv.lock (PR lock minus pytest-shard, which our main drops) - Dockerfile: make z3-solver libz3 discoverable for fla/tilelang - modelopt worker: register "nemo_rl." as an mbridge allowed target prefix (new API in this Megatron-Bridge) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Signed-off-by: Zhaopeng Qiu <alexq@nvidia.com>
What does this PR do ?
Add a one line overview of what this PR aims to accomplish.
Issues
List issues that this PR closes (syntax):
Usage
# Add a code snippet demonstrating how to use thisBefore your PR is "Ready for review"
Pre checks:
Additional Information