Allow CUDA VMM feature transport with the Rust frontend (#39347)

This commit is contained in:
Lianmin Zheng
2026-09-14 17:09:49 -07:00
committed by GitHub
parent 6e755e4114
commit a2b4e8888f
2 changed files with 6 additions and 7 deletions
@@ -851,11 +851,6 @@ def handle_multimodal_feature_transport(server_args: Any):
raise ValueError(
"--mm-feature-transport=cuda_vmm does not support pipeline parallelism."
)
if envs.SGLANG_RUST_SERVER.get():
raise ValueError(
"--mm-feature-transport=cuda_vmm is not supported with "
"SGLANG_RUST_SERVER."
)
pool_budget_mb = envs.SGLANG_MM_FEATURE_CACHE_MB.get()
handle_kind = "CUDA FABRIC" if cfg.nnodes > 1 else "POSIX FD"
logger.info(