Skip to content

fix: Add model registration to SGLang multimodal workers fixing bug #4486 #4512

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 3 commits into
base: main
Choose a base branch
from
Open

fix: Add model registration to SGLang multimodal workers fixing bug #4486 #4512

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
5a6e6d9
fix: Add model registration to SGLang multimodal workers
nancya-nv Nov 19, 2025
232bf97
Merge branch 'main' into fix/sglang-multimodal-worker-registration
nancya-nv Nov 20, 2025
44a18bd
Merge branch 'main' into fix/sglang-multimodal-worker-registration
krishung5 Nov 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
55 changes: 42 additions & 13 deletions components/src/dynamo/sglang/main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -376,16 +376,24 @@ async def init_multimodal_encode_worker(runtime: DistributedRuntime, config: Con

await pd_worker_client.wait_for_instances()

tasks = [
generate_endpoint.serve_endpoint(
handler.generate,
graceful_shutdown=True,
metrics_labels=[("model", server_args.served_model_name)],
)
]
ready_event = asyncio.Event()

try:
await asyncio.gather(*tasks)
await asyncio.gather(
generate_endpoint.serve_endpoint(
handler.generate,
graceful_shutdown=True,
metrics_labels=[("model", server_args.served_model_name)],
),
register_llm_with_readiness_gate(
None, # encode worker doesn't have engine
generate_endpoint,
server_args,
dynamo_args,
input_type=ModelInput.Text,
readiness_gate=ready_event,
),
)
except Exception as e:
logging.error(f"Failed to serve endpoints: {e}")
raise
Expand Down Expand Up @@ -419,11 +427,24 @@ async def init_multimodal_worker(runtime: DistributedRuntime, config: Config):

await handler.async_init()

health_check_payload = SglangHealthCheckPayload(engine).to_dict()
ready_event = asyncio.Event()

try:
await generate_endpoint.serve_endpoint(
handler.generate,
metrics_labels=[("model", server_args.served_model_name)],
graceful_shutdown=True,
await asyncio.gather(
generate_endpoint.serve_endpoint(
handler.generate,
metrics_labels=[("model", server_args.served_model_name)],
graceful_shutdown=True,
health_check_payload=health_check_payload,
),
register_llm_with_readiness_gate(
engine,
generate_endpoint,
server_args,
dynamo_args,
readiness_gate=ready_event,
),
)
except Exception as e:
logging.error(f"Failed to serve endpoints: {e}")
Expand All @@ -448,6 +469,7 @@ async def init_multimodal_prefill_worker(runtime: DistributedRuntime, config: Co
await handler.async_init()

health_check_payload = SglangPrefillHealthCheckPayload(engine).to_dict()
ready_event = asyncio.Event()

try:
await asyncio.gather(
Expand All @@ -456,7 +478,14 @@ async def init_multimodal_prefill_worker(runtime: DistributedRuntime, config: Co
graceful_shutdown=True,
metrics_labels=[("model", server_args.served_model_name)],
health_check_payload=health_check_payload,
)
),
register_llm_with_readiness_gate(
engine,
generate_endpoint,
server_args,
dynamo_args,
readiness_gate=ready_event,
),
)
except Exception as e:
logging.error(f"Failed to serve endpoints: {e}")
Expand Down
Loading