A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 7, 10:15 UTC
[Misc] Fix typos in comments: explict→explicit, paramaters→parameters (#35648)
lin-shh committed
Mar 3, 2026 at 00:59 UTC
35a6f0bfe2c2b9e2c7220a3307aaa472b219726e
3 files changed
+3
-3
vllm/model_executor/layers/fused_moe/oracle/fp8.py
+1
-1
@@ -603,7 +603,7 @@ def make_fp8_moe_kernel(
603
)
604
605
# NOTE(rob): we only want the mk to control the shared_expert
606
- # if using all2all (for SBO). bnell is making this explict in
606
+ # if using all2all (for SBO). bnell is making this explicit in
607
# the new MoE runner class.
608
kernel = mk.FusedMoEModularKernel(
609
prepare_finalize,
vllm/model_executor/layers/fused_moe/oracle/nvfp4.py
+1
-1
@@ -458,7 +458,7 @@ def make_nvfp4_moe_kernel(
458
)
459
460
# NOTE(rob): we only want the mk to control the shared_expert
461
- # if using all2all (for SBO). bnell is making this explict in
461
+ # if using all2all (for SBO). bnell is making this explicit in
462
# the new MoE runner class.
463
kernel = mk.FusedMoEModularKernel(
464
prepare_finalize,
vllm/model_executor/layers/fused_moe/router/router_factory.py
+1
-1
@@ -44,7 +44,7 @@ def create_fused_moe_router(
44
# grouped topk + fused topk bias parameters
45
routed_scaling_factor: float = 1.0,
46
e_score_correction_bias: torch.Tensor | None = None,
47
- # custom routing paramaters
47
+ # custom routing parameters
48
custom_routing_function: Callable | None = None,
49
# eplb parameters
50
enable_eplb: bool = False,