Ensure that inference dummy_forward does not try to match on a cudagraph when running eager#3815
Conversation
|
This PR has been automatically converted to draft because all PRs must start as drafts. When you are ready for review, click Ready for Review to begin the review process. This will:
See the contribution guide for more details. |
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
| # cudagraph check (which gates on using_cuda_graph_this_step()) instead of | ||
| # the training-mode check (which unconditionally replays captured graphs | ||
| # when inference_context is not in kwargs). | ||
| return self.model( |
There was a problem hiding this comment.
I'm a bit scared that this will lead to segfaults. If we pass the inference context, we will attempt to write into the kv-cache/mamba-cache. But we are not setting up the required pointers properly.
There was a problem hiding this comment.
Removed this in favour of gating on inference_mode instead
svcnvidia-nemo-ci
added
Approved
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/23060422351 |
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/23062146973 |
7 participants
What does this PR do ?
Fix for #3789.
Contribution process
Pre-checks
Code review
Feel free to message or comment the @mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!
All PRs start as draft. If you open a non-draft PR, it will be automatically converted to draft.
Step 1: Mark PR as "Ready for Review"
.github/CODEOWNERS.Final Review might get declined if these requirements are not fulfilled.
Step 2: Final Review
For PRs that change
megatron/core, once all expert reviewers have approved, theFinal Reviewlabel is applied automatically and final reviewers are assigned.For PRs outside
megatron/core, this step is skipped.Step 3: Approved
Once all required reviewers have approved, the
Approvedlabel is applied automatically.Merge
Any member of mcore-engineers will be able to merge your PR.
For MRs into `dev` branch
The proposed review process for `dev` branch is under active discussion.MRs are mergable after one approval by either
eharper@nvidia.comorzijiey@nvidia.com.