biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Feat][RL][1/2] Native Weight Syncing API: NCCL (#31943)

Browse Source 
Signed-off-by: ahao-anyscale <ahao@anyscale.com>
Signed-off-by: Aaron Hao <ahao@anyscale.com>
Co-authored-by: SumanthRH <sumanthrh99@gmail.com>
This commit is contained in:
Aaron Hao
2026-02-05 09:13:23 -08:00
committed by GitHub
parent 82914d2ae8
commit c1858b7ec8
27 changed files with 2974 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
.buildkite/test_areas/distributed.yaml
View File

@@ -62,6 +62,7 @@ steps:
- tests/compile/fullgraph/test_basic_correctness.py
- examples/offline_inference/rlhf.py
- examples/offline_inference/rlhf_colocate.py
- examples/offline_inference/new_weight_syncing/
- tests/examples/offline_inference/data_parallel.py
- tests/v1/distributed
- tests/v1/engine/test_engine_core_client.py
@@ -96,9 +97,14 @@ steps:
- pytest -v -s distributed/test_symm_mem_allreduce.py
# TODO: create a dedicated test section for multi-GPU example tests
# when we have multiple distributed example tests
# OLD rlhf examples
- cd ../examples/offline_inference
- VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 rlhf.py
- VLLM_ALLOW_INSECURE_SERIALIZATION=1 RAY_DEDUP_LOGS=0 python3 rlhf_colocate.py
# NEW rlhf examples
- cd new_weight_syncing
- VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 rlhf.py
- VLLM_ALLOW_INSECURE_SERIALIZATION=1 python3 rlhf_async_new_apis.py
- label: Distributed Tests (8 GPUs)(H100)
timeout_in_minutes: 10