Repository navigation
feat: fix the vLLM DP path - #2517
Conversation
Signed-off-by: Guyue Huang <guyueh@login-lyris01.lyris.clusters.nvidia.com>
|
/ok to test efc6fc2 |
Signed-off-by: Guyue Huang <guyueh@login-lyris01.lyris.clusters.nvidia.com>
|
/ok to test 9f381f2 |
|
fast CI is failing for |
|
/ok to test 312e0ce |
Signed-off-by: Alan Liu <alanliu@nvidia.com>
|
/ok to test f38d77a |
@alanliu-nvidia, there was an error processing your request: See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/2/ |
|
/ok to test f38d77a |
|
Thanks for sharing the |
|
GB200 FP8 refit follow-up (sync and async): both 4n4g jobs completed 3/3 GRPO steps with repeated weight refits and Slurm exit |
|
Current-head GB200 validation update (SHA |
Signed-off-by: Alan Liu <alanliu@nvidia.com>
|
/ok to test c465781 |
|
CI follow-up: the prior L2 run ( |
Signed-off-by: Alan Liu <alanliu@nvidia.com>
|
/ok to test a6c8c3a |
|
@andresnowak Thanks for the |
|
CI update for current head |
|
/ok to test 4e10de3 |
Signed-off-by: Alan Liu <alanliu@nvidia.com>
|
/ok to test 8610bbe |
|
CI unit-test fix follow-up: the two native-DP empty-shard test fakes now initialize |
|
/ok to test dbed9e4 |
What does this PR do ?
Previously nemo-rl doesn't work for vllm's native DP (EP>TP), this PR wants to support this case.
The following basic tests have passed, now trying the nightly test
New nightly test figures:

H100 with EP=8 async engine
https://wandb.ai/nvidia/nemo-rl/runs/4mcplb63
H100 with TP=2 EP=16 sync engine

https://wandb.ai/nvidia/nemo-rl/runs/b3mon8zg
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
Closes #1101
FP8 MoE refit scope
FP8 MoE refit scope: FlashInfer TRT-LLM FP8 MoE refit is not supported by this PR. Non-DeepSeek-V4 layouts that change weight shape or dtype fail explicitly rather than proceeding with an unvalidated refit. The Moonlight recipe uses
moe_backend: triton; no FlashInfer TRT-LLM FP8 MoE refit GPU validation is claimed.