Update #2612#2879
Draft
tdene wants to merge 22 commits into
Draft
Conversation
Hand-port of the topology-aware actor placement feature, combining the net effect of upstream commits 8bd2417 (NVLink-aware training) and 2c2b9f60 (topology-aware inference placement). The topology logic is grafted onto the existing setup shape directly so it doesn't carry along the surrounding NeMo Gym reservation block from the source branch. virtual_cluster.py: add NVLINK_DOMAIN_*/TOPO_RANK_* constants and DEFAULT_PORT_RANGE_*, get_ray_cluster_topology(), select_segment_nodes(), _sort_bundle_indices_by_topology(); replace GetGPUIDActor with a _get_gpu_id_info Ray task that also returns (nvlink_domain, topo_rank); add port_range_low/high, segment_size, node_resource_constraints params and _nvlink_domain_per_bundle_index state to RayVirtualCluster; merge node_resource_constraints into bundle specs; topology-aware _get_sorted_bundle_indices. vllm_generation.py: add init_cluster_placement_groups staticmethod for deterministic PG ordering when other components compete for Ray resources; add topology arguments to allocate_worker_groups and warn when a model-parallel group straddles NVLink domains. grpo.py: read cluster.segment_size, build node_resource_constraints from get_ray_cluster_topology()/select_segment_nodes() in non-colocated setup, relocate inference cluster creation, call VllmGeneration.init_cluster_placement_groups so inference PGs claim domain-aligned nodes first. ray.sub: write topology_probe.sh that parses ClusterUUID from nvidia-smi -q and topo_rank from SLURM_TOPOLOGY_ADDR, source it before each ray start, and register nvlink_domain_<uuid>/topo_rank as Ray resources. Added \`export RAY_RESOURCES\` (missing in upstream 8bd2417), required so the variable propagates to the \`bash /launch-head.sh\` child invocation. Signed-off-by: Youngeun Kwon <youngeunk@nvidia.com>
Signed-off-by: Youngeun Kwon <youngeunk@nvidia.com>
Signed-off-by: Youngeun Kwon <youngeunk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
…JSON Signed-off-by: Terry Kong <terryk@nvidia.com>
This reverts commit 2c7af4e. Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
…erplate Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
…rence topology Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
…isting configs Signed-off-by: Terry Kong <terryk@nvidia.com>
Signed-off-by: Terry Kong <terryk@nvidia.com>
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Teodor-Dumitru Ene <teodord.ene@gmail.com>
Signed-off-by: Teodor-Dumitru Ene <teodord.ene@gmail.com>
github-actions
Bot
added
Documentation
✅ Submodule Fast-Forward Check ResultsCheck based on commit: 48c74fa (PR #2879 from ✅ Submodules that are properly updated:Automodel: ✅ PR branch is ahead of youngeunk/topology-aware-placement branch (fast-forward) All submodule changes look good! ✨ |
terrykong
force-pushed
the
youngeunk/topology-aware-placement
branch
2 times, most recently
from
0f2632b to
6874bf5
Compare
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
4 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.
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