vllm.v1.worker.gpu ¶
Modules:
-
async_utils– -
cp_utils– -
cudagraph_utils– -
dp_utils– -
eplb_utils– -
kv_connector– -
mm– -
model_runner–NOTE: Coding style guide for this file:
-
model_states– -
pool– -
pp_utils–Pipeline Parallelism utils for V2 Model Runner.
-
sample– -
spec_decode– -
states– -
warmup–