[NV] DSR1 FP8 GB300 Dynamo TRT (8k1k and 1k1k) - #627
Merged
Conversation
Contributor
|
Claude finished @jthomson04's task in 1m 15s —— View job PR Review Complete
SummaryThis PR adds DeepSeek R1 FP8 GB300 Dynamo TRT-LLM disaggregated multinode configurations for 8k1k and 1k1k sequence lengths. Found 1 blocking issue: 🔴 Model path uses local Validations passed:
|
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Signed-off-by: jthomson04 <jothomson@nvidia.com>
Signed-off-by: jthomson04 <jothomson@nvidia.com>
Signed-off-by: jthomson04 <jothomson@nvidia.com>
Signed-off-by: jthomson04 <jothomson@nvidia.com>
jthomson04
force-pushed
the
nv/dsr1-fp8-gb300-dynamo-trt
branch
from
February 4, 2026 18:43
9acc0e8 to
504fd22
Compare
Signed-off-by: jthomson04 <jothomson@nvidia.com>
functionstackx
requested changes
Feb 5, 2026
Signed-off-by: jthomson04 <jothomson@nvidia.com>
This was referenced Feb 5, 2026
Collaborator
|
Successful sweep - https://github.com/InferenceMAX/InferenceMAX/actions/runs/21697417250 |
Collaborator
|
@claude In the .github/configs/nvidia-master.yaml file, please go through all added downstream config files (the links) and ensure there is no system specific config settings or hardcoded paths to cache artifacts, etc. |
Collaborator
|
Claude finished @cquil11's task in 2m 8s —— View job Reviewing Downstream Config Files for System-Specific Settings
SummaryReviewed all 27 downstream config files from the ✅ No Blocking Issues FoundGood news: None of the downstream config files contain hardcoded system paths like
Environment-Specific Configurations (Expected/Acceptable)The configs do contain hardware-specific settings that are expected and appropriate for this use case:
These are not problematic - they are intentional hardware-specific optimizations that define the benchmark configuration. Files Reviewed1k1k MTP configs (7 files): ✅ Clean
1k1k STP configs (7 files): ✅ Clean
8k1k MTP configs (6 files): ✅ Clean
8k1k STP configs (7 files): ✅ Clean
|
Collaborator
|
@claude Please update the PR description in accordance with changes introduced in this PR, as well as changes stated in perf-changelog.yaml. |
Collaborator
|
Claude finished @cquil11's task in 1m 0s —— View job Updating PR Description
SummaryUpdated the PR description with a comprehensive summary based on the changes introduced: Key changes documented:
|
Oseltamivir
added a commit
that referenced
this pull request
Aug 5, 2026
…lamp Fixes the defect instead of stepping around it. b368eae clamped the low-latency ladder to 128 so the corrupt 256 rung was not measured; the corruption itself was still in the kernels we publish numbers from, at roughly 1.5-3.3% per invocation on every Blackwell SKU. The fix was already upstream. DeepEP PR #642 adds a CTA-scope fence.proxy.async.shared::cta before mbarrier_arrive(empty_barriers[stage_idx]) in LOW_LATENCY_COMBINE_RECV, so the combine consumer's shared-memory reads retire before the stage is declared empty and the producer's next TMA load refills it -- signalling empty too early is what let one output row be assembled from two tiles, which matches the observed signature exactly (norm preserved to 4 s.f., 16-40% of elements wrong). It closed #621, the same race found independently on NVL72. COLLX_DEEPEP_V2_COMMIT was fa8a9b16, the head of the pre-merge PR #605 branch, cut before #642 landed. Its one unique commit was the #630 single-node V2 init fix, which main carries as 56169594e, so moving to main loses nothing and also picks up #715 (system-scope release before the GIN barrier when scale-up spans NVLink and RDMA), #688 (NCCL Device API: runtime version for ncclDevCommCreate), #178 (SM90), #641, and #640/#627 upstream's own libnccl and SO-name handling. Verified before bumping that main still exposes every API this adapter calls: legacy Buffer kwargs incl. allow_mnnvl, get_low_latency_rdma_size_hint, low_latency_dispatch/combine, ElasticBuffer's full kwarg set, topk_idx_t, and the fp8 cast helpers. rewrite_deepep_v2 now succeeds when the source already matches, because main contains #640 and the old count(old) == 1 assertion would have aborted every leg at repository-stage. Verified against all four source states: old form rewrites, fixed form is a no-op, absent and duplicated forms still fail. _LL_LADDER_CAP returns to 256. The two constants stay separate: the receive must not be sized from max(ladder), or clamping the ladder would change the footprint that drives transport traffic and fp8 dequant volume. The cap test now asserts ladder <= buffer -- strict inequality encoded the workaround, not an invariant. The backend cache key includes the pin, so this forces a rebuild rather than reusing the old build.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This PR adds DeepSeek R1 FP8 GB300 Dynamo TRT-LLM disaggregated multinode configurations for 8k1k and 1k1k sequence lengths.
Changes
New Configuration:
dsr1-fp8-gb300-dynamo-trtnvcr.io/nvidia/ai-dynamo/tensorrtllm-runtime:0.8.1.post2deepseek-ai/DeepSeek-R1-0528Sequence Length Configurations
1k1k (ISL=1024, OSL=1024)
8k1k (ISL=8192, OSL=1024)
Launch Script Updates (
runners/launch_gb300-nv.sh)/raid/shared/models/deepseek-r1-0528-fp4-v2/raid/shared/models/deepseek-r1-0528SRT_SLURM_MODEL_PREFIXvariable to correctly map model aliases in srt-slurmBenchmark Verification
Related PRs