Skip to content

Fp8 debug - #2

Draft
wdykas wants to merge 82 commits into
mainfrom
fp8-debug
Draft

Fp8 debug#2
wdykas wants to merge 82 commits into
mainfrom
fp8-debug

Conversation

@wdykas

@wdykas wdykas commented Mar 6, 2026

Copy link
Copy Markdown
Owner

What does this PR do ?

⚠️ For major changes (either in lines of code or in its impact), please make sure to first share a design doc with the team. If you're unsure what's the best way to do so, contact the @mcore-oncall.

Contribution process

flowchart LR
    A[Pre-checks] --> B[PR Tests]
    subgraph Code Review/Approval
        C1[Expert Review] --> C2[Final Review]
    end
    B --> C1
    C2 --> D[Merge]
Loading

Pre-checks

  • I want this PR in a versioned release and have added the appropriate Milestone (e.g., Core 0.8)
  • I have added relevant unit tests
  • I have added relevant functional tests
  • I have added proper typing to my code Typing guidelines
  • I have added relevant documentation
  • I have run the autoformatter.sh on my PR

Code review

The following process is enforced via the CODEOWNERS file for changes into megatron/core. For changes outside of megatron/core, it is up to the PR author whether or not to tag the Final Reviewer team.

For MRs into `main` branch

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!

(Step 1): Add PR label Expert Review

(Step 2): Collect the expert reviewers reviews

  1. Attach the Expert Review label when your PR is ready for review.
  2. GitHub auto-assigns expert reviewers based on your changes. They will get notified and pick up your PR soon.

⚠️ Only proceed to the next step once all reviewers have approved, merge-conflict are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

(Step 3): Final Review

  1. Add Final Review label
  2. GitHub auto-assigns final reviewers based on your changes. They will get notified and pick up your PR soon.

(Optional Step 4): Cherry-pick into release branch

If this PR also needs to be merged into core_r* release branches, after this PR has been merged, select Cherry-pick to open a new PR into the release branch.

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.com or zijiey@nvidia.com.

Merging your PR

Any member of core-adlr and core-nemo will be able to merge your PR.

ko3n1g and others added 28 commits March 6, 2026 15:39
Signed-off-by: oliver könig <okoenig@nvidia.com>
…`. (NVIDIA#2639)

Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
Signed-off-by: Keshav Santhanam <ksanthanam@nvidia.com>
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
Co-authored-by: Eric Harper <complex451@gmail.com>
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
Co-authored-by: Philip Petrakian <ppetrakian@nvidia.com>
Signed-off-by: dimapihtar <dpihtar@gmail.com>
Co-authored-by: Eric Harper <eharper@nvidia.com>
wdykas added a commit that referenced this pull request Jun 2, 2026
Validated the NVSHMEM data path end-to-end on 2x GB200: 4 KV blobs
through a depth-2 credit ring (PE0->PE1) arrive intact and in order,
which means the ack-credit path recycled twice (send #2 blocked until
decode acked #0) -- decode-gating, ring-wrap, and one-way put_signal
all confirmed on the real interconnect, not just the simulator.

Adds a guarded pytest (skips unless nvshmem4py + >=2 CUDA devices) so
the hardware path is reproducible. CPU CI still covers the protocol via
the in-memory simulator (test_nvshmem_ring.py); NCCL remains the
default. Tested with nvidia-nvshmem-cu12 / nvshmem4py-cu12 3.6.5.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
wdykas pushed a commit that referenced this pull request Jun 4, 2026
* mamba_mixer.py: move BIK decode kernel + buffer helpers to
  megatron/core/ssm/ops/bik_decode.py (santhnm2 NVIDIA#4). The mixer now
  delegates to module-level functions and only owns the buffer
  dataclass instance.
* mamba_mixer.py: clarify drift comment — note the ~6e-5 / ~2.4e-4
  numbers are empirical (test_mamba_decode.py, nemotron6_3b_moe dims)
  rather than analytic (santhnm2 #1, NVIDIA#11).
* mamba_mixer.py: replace the Claude-injected unicode "buffer ⊕ new_tok"
  with plain "buffer + new_tok" in two docstrings (santhnm2 #2).
* chat_completions.py: remove the local BatchEncoding shim — already
  fixed upstream in !4732 (santhnm2 NVIDIA#5).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.