[Feature] Support recording expert indices for rollout router replay (#28284)

Signed-off-by: xhx1022 <1737006628@qq.com>
Signed-off-by: Hongxin Xu <70438206+xhx1022@users.noreply.github.com>
Signed-off-by: arlenxu <arlenxu@tencent.com>
Co-authored-by: 22quinn <33176974+22quinn@users.noreply.github.com>
Co-authored-by: arlenxu <arlenxu@tencent.com>
This commit is contained in:
Hongxin Xu
2026-01-12 06:23:04 -08:00
committed by GitHub
co-authored by 22quinn arlenxu
parent 0565f1fdec
commit 49e6b86c91
11 changed files with 463 additions and 3 deletions
+6
View File
@@ -354,6 +354,7 @@ class EngineArgs:
"""Arguments for vLLM engine."""
model: str = ModelConfig.model
enable_return_routed_experts: bool = ModelConfig.enable_return_routed_experts
model_weights: str = ModelConfig.model_weights
served_model_name: str | list[str] | None = ModelConfig.served_model_name
tokenizer: str | None = ModelConfig.tokenizer
@@ -657,6 +658,10 @@ class EngineArgs:
**model_kwargs["allow_deprecated_quantization"],
)
model_group.add_argument("--enforce-eager", **model_kwargs["enforce_eager"])
model_group.add_argument(
"--enable-return-routed-experts",
**model_kwargs["enable_return_routed_experts"],
)
model_group.add_argument("--max-logprobs", **model_kwargs["max_logprobs"])
model_group.add_argument("--logprobs-mode", **model_kwargs["logprobs_mode"])
model_group.add_argument(
@@ -1239,6 +1244,7 @@ class EngineArgs:
quantization=self.quantization,
allow_deprecated_quantization=self.allow_deprecated_quantization,
enforce_eager=self.enforce_eager,
enable_return_routed_experts=self.enable_return_routed_experts,
max_logprobs=self.max_logprobs,
logprobs_mode=self.logprobs_mode,
disable_sliding_window=self.disable_sliding_window,