[RL] Add sharding-aware NCCL M2N weight transfer (#51520)

Signed-off-by: Ke Wen <kwen@nvidia.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Co-authored-by: aoshen02 <aoshen@inferact.ai>
This commit is contained in:
Ke Wen
2026-09-26 01:38:53 +00:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> aoshen02
parent 4bb804cc5b
commit b6d2b4fc25
13 changed files with 2812 additions and 6 deletions
+3 -1
View File
@@ -35,6 +35,7 @@ engine drives on your behalf:
| [IPC](ipc.md) | CUDA IPC handles | Colocated training and inference on same GPU |
| [sparse_nccl](nccl.md#sparse-nccl) | NCCL broadcast | Checkpoint-coordinate sparse weight patches |
| [sharded_rdt](sharded_rdt.md) | NIXL / Ray Direct Transport (pull-based) | Very large models where each worker needs only its own slice (MoE with expert parallelism) |
| [nccl_m2n](m2n.md) | NCCL M2N reshard | Trainer and inference use different sharding layouts |
## Quickstart
@@ -49,7 +50,8 @@ from vllm.config import WeightTransferConfig
llm = LLM(
model="my-model",
weight_transfer_config=WeightTransferConfig(backend="nccl"), # or "ipc", "sparse_nccl", "sharded_rdt"
# Other backends: "ipc", "sparse_nccl", "sharded_rdt", "nccl_m2n".
weight_transfer_config=WeightTransferConfig(backend="nccl"),
)
```